Datasets:

jamescalam

/

llama-2-arxiv-papers-chunked

like 17

1409.0473

9

sentence (x1;x2;:::;x T).In a new model architecture, we define each conditional probability in Eq. (2) as: p(yijy1;:::;y i1;x) =g(yi1;si;ci); (4) wheresiis an RNN hidden state for time i, computed by si=f(si1;yi1;ci): It should be noted that unlike the existing encoder–decoder approach (see Eq. (2)), here the probability is conditioned on a distinct context vector cifor each target word yi. The context vector cidepends on a sequence of annotations (h1;


;hTx)to which an encoder maps the input sentence. Each annotationhicontains information about the whole input sequence with a strong focus on the parts surrounding the i-th word of the input sequence. We explain in detail how the annotations are computed in the next section. The context vector ciis, then, computed as a weighted sum of these annotationshi: ci=TxX j=1ijhj: (5) The weightijof each annotation hjis computed by ij=exp (eij)PTx

1409.0473

Neural Machine Translation by Jointly Learning to Align and Translate

Neural machine translation is a recently proposed approach to machine translation. Unlike the traditional statistical machine translation, the neural machine translation aims at building a single neural network that can be jointly tuned to maximize the translation performance. The models proposed recently for neural machine translation often belong to a family of encoder-decoders and consists of an encoder that encodes a source sentence into a fixed-length vector from which a decoder generates a translation. In this paper, we conjecture that the use of a fixed-length vector is a bottleneck in improving the performance of this basic encoder-decoder architecture, and propose to extend this by allowing a model to automatically (soft-)search for parts of a source sentence that are relevant to predicting a target word, without having to form these parts as a hard segment explicitly. With this new approach, we achieve a translation performance comparable to the existing state-of-the-art phrase-based system on the task of English-to-French translation. Furthermore, qualitative analysis reveals that the (soft-)alignments found by the model agree well with our intuition.

http://arxiv.org/pdf/1409.0473

Accepted at ICLR 2015 as oral presentation

cs.CL

20140901

20160519

1409.0473

10

ci=TxX j=1ijhj: (5) The weightijof each annotation hjis computed by ij=exp (eij)PTx k=1exp (eik); (6) where eij=a(si1;hj) is an alignment model which scores how well the inputs around position jand the output at position imatch. The score is based on the RNN hidden state si1(just before emitting yi, Eq. (4)) and the j-th annotation hjof the input sentence. We parametrize the alignment model aas a feedforward neural network which is jointly trained with all the other components of the proposed system. Note that unlike in traditional machine translation, 3 Published as a conference paper at ICLR 2015 the alignment is not considered to be a latent variable. Instead, the alignment model directly computes a soft alignment, which allows the gradient of the cost function to be backpropagated through. This gradient can be used to train the alignment model as well as the whole translation model jointly. We can understand the approach of taking a weighted sum of all the annotations as computing an expected annotation , where the expectation is over possible alignments. Let ijbe a probability that

1409.0473

Neural Machine Translation by Jointly Learning to Align and Translate

Neural machine translation is a recently proposed approach to machine translation. Unlike the traditional statistical machine translation, the neural machine translation aims at building a single neural network that can be jointly tuned to maximize the translation performance. The models proposed recently for neural machine translation often belong to a family of encoder-decoders and consists of an encoder that encodes a source sentence into a fixed-length vector from which a decoder generates a translation. In this paper, we conjecture that the use of a fixed-length vector is a bottleneck in improving the performance of this basic encoder-decoder architecture, and propose to extend this by allowing a model to automatically (soft-)search for parts of a source sentence that are relevant to predicting a target word, without having to form these parts as a hard segment explicitly. With this new approach, we achieve a translation performance comparable to the existing state-of-the-art phrase-based system on the task of English-to-French translation. Furthermore, qualitative analysis reveals that the (soft-)alignments found by the model agree well with our intuition.

http://arxiv.org/pdf/1409.0473

Accepted at ICLR 2015 as oral presentation

cs.CL

20140901

20160519

1409.0473

11

We can understand the approach of taking a weighted sum of all the annotations as computing an expected annotation , where the expectation is over possible alignments. Let ijbe a probability that the target word yiis aligned to, or translated from, a source word xj. Then, thei-th context vector ciis the expected annotation over all the annotations with probabilities ij. The probability ij, or its associated energy eij, reflects the importance of the annotation hjwith respect to the previous hidden state si1in deciding the next state siand generating yi. Intuitively, this implements a mechanism of attention in the decoder. The decoder decides parts of the source sentence to pay attention to. By letting the decoder have an attention mechanism, we relieve the encoder from the burden of having to encode all information in the source sentence into a fixedlength vector. With this new approach the information can be spread throughout the sequence of annotations, which can be selectively retrieved by the decoder accordingly. 3.2 E NCODER : BIDIRECTIONAL RNN FOR ANNOTATING SEQUENCES The usual RNN, described in Eq. (1), reads an input sequence xin order starting from the first

1409.0473

Neural Machine Translation by Jointly Learning to Align and Translate

Neural machine translation is a recently proposed approach to machine translation. Unlike the traditional statistical machine translation, the neural machine translation aims at building a single neural network that can be jointly tuned to maximize the translation performance. The models proposed recently for neural machine translation often belong to a family of encoder-decoders and consists of an encoder that encodes a source sentence into a fixed-length vector from which a decoder generates a translation. In this paper, we conjecture that the use of a fixed-length vector is a bottleneck in improving the performance of this basic encoder-decoder architecture, and propose to extend this by allowing a model to automatically (soft-)search for parts of a source sentence that are relevant to predicting a target word, without having to form these parts as a hard segment explicitly. With this new approach, we achieve a translation performance comparable to the existing state-of-the-art phrase-based system on the task of English-to-French translation. Furthermore, qualitative analysis reveals that the (soft-)alignments found by the model agree well with our intuition.

http://arxiv.org/pdf/1409.0473

Accepted at ICLR 2015 as oral presentation

cs.CL

20140901

20160519

1409.0473

12

The usual RNN, described in Eq. (1), reads an input sequence xin order starting from the first symbolx1to the last one xTx. However, in the proposed scheme, we would like the annotation of each word to summarize not only the preceding words, but also the following words. Hence, we propose to use a bidirectional RNN (BiRNN, Schuster and Paliwal, 1997), which has been successfully used recently in speech recognition (see, e.g., Graves et al. , 2013). A BiRNN consists of forward and backward RNN’s. The forward RNN !freads the input sequence as it is ordered (from x1toxTx) and calculates a sequence of forward hidden states ( !h1;


; !hTx). The backward RNN freads the sequence in the reverse order (from xTxtox1), resulting in a sequence of backward hidden states ( h1;


; hTx). We obtain an annotation for each word xjby concatenating the forward hidden state !hjand the backward one hj, i.e.,hj=h !h> j; h> ji>

1409.0473

Neural Machine Translation by Jointly Learning to Align and Translate

Neural machine translation is a recently proposed approach to machine translation. Unlike the traditional statistical machine translation, the neural machine translation aims at building a single neural network that can be jointly tuned to maximize the translation performance. The models proposed recently for neural machine translation often belong to a family of encoder-decoders and consists of an encoder that encodes a source sentence into a fixed-length vector from which a decoder generates a translation. In this paper, we conjecture that the use of a fixed-length vector is a bottleneck in improving the performance of this basic encoder-decoder architecture, and propose to extend this by allowing a model to automatically (soft-)search for parts of a source sentence that are relevant to predicting a target word, without having to form these parts as a hard segment explicitly. With this new approach, we achieve a translation performance comparable to the existing state-of-the-art phrase-based system on the task of English-to-French translation. Furthermore, qualitative analysis reveals that the (soft-)alignments found by the model agree well with our intuition.

http://arxiv.org/pdf/1409.0473

Accepted at ICLR 2015 as oral presentation

cs.CL

20140901

20160519

1409.0473

13

backward one hj, i.e.,hj=h !h> j; h> ji> . In this way, the annotation hjcontains the summaries of both the preceding words and the following words. Due to the tendency of RNNs to better represent recent inputs, the annotation hjwill be focused on the words around xj. This sequence of annotations is used by the decoder and the alignment model later to compute the context vector (Eqs. (5)–(6)). See Fig. 1 for the graphical illustration of the proposed model. 4 E XPERIMENT SETTINGS We evaluate the proposed approach on the task of English-to-French translation. We use the bilingual, parallel corpora provided by ACL WMT ’14.3As a comparison, we also report the performance of an RNN Encoder–Decoder which was proposed recently by Cho et al. (2014a). We use the same training procedures and the same dataset for both models.4 4.1 D ATASET WMT ’14 contains the following English-French parallel corpora: Europarl (61M words), news commentary (5.5M), UN (421M) and two crawled corpora of 90M and 272.5M words respectively,

1409.0473

Neural Machine Translation by Jointly Learning to Align and Translate

Neural machine translation is a recently proposed approach to machine translation. Unlike the traditional statistical machine translation, the neural machine translation aims at building a single neural network that can be jointly tuned to maximize the translation performance. The models proposed recently for neural machine translation often belong to a family of encoder-decoders and consists of an encoder that encodes a source sentence into a fixed-length vector from which a decoder generates a translation. In this paper, we conjecture that the use of a fixed-length vector is a bottleneck in improving the performance of this basic encoder-decoder architecture, and propose to extend this by allowing a model to automatically (soft-)search for parts of a source sentence that are relevant to predicting a target word, without having to form these parts as a hard segment explicitly. With this new approach, we achieve a translation performance comparable to the existing state-of-the-art phrase-based system on the task of English-to-French translation. Furthermore, qualitative analysis reveals that the (soft-)alignments found by the model agree well with our intuition.

http://arxiv.org/pdf/1409.0473

Accepted at ICLR 2015 as oral presentation

cs.CL

20140901

20160519

1409.0473

14

commentary (5.5M), UN (421M) and two crawled corpora of 90M and 272.5M words respectively, totaling 850M words. Following the procedure described in Cho et al. (2014a), we reduce the size of the combined corpus to have 348M words using the data selection method by Axelrod et al. (2011).5 We do not use any monolingual data other than the mentioned parallel corpora, although it may be possible to use a much larger monolingual corpus to pretrain an encoder. We concatenate news-test3http://www.statmt.org/wmt14/translation-task.html 4Implementations are available at https://github.com/lisa-groundhog/GroundHog . 5Available online at http://www-lium.univ-lemans.fr/ ˜schwenk/cslm_joint_paper/ . 4 Published as a conference paper at ICLR 2015 0 10 20 30 40 50 60 Sentence length051015202530BLEU score RNNsearch-50 RNNsearch-30 RNNenc-50 RNNenc-30 Figure 2: The BLEU scores of the generated translations on the test set with respect to the lengths of the sentences. The results are on

1409.0473

Neural Machine Translation by Jointly Learning to Align and Translate

Neural machine translation is a recently proposed approach to machine translation. Unlike the traditional statistical machine translation, the neural machine translation aims at building a single neural network that can be jointly tuned to maximize the translation performance. The models proposed recently for neural machine translation often belong to a family of encoder-decoders and consists of an encoder that encodes a source sentence into a fixed-length vector from which a decoder generates a translation. In this paper, we conjecture that the use of a fixed-length vector is a bottleneck in improving the performance of this basic encoder-decoder architecture, and propose to extend this by allowing a model to automatically (soft-)search for parts of a source sentence that are relevant to predicting a target word, without having to form these parts as a hard segment explicitly. With this new approach, we achieve a translation performance comparable to the existing state-of-the-art phrase-based system on the task of English-to-French translation. Furthermore, qualitative analysis reveals that the (soft-)alignments found by the model agree well with our intuition.

http://arxiv.org/pdf/1409.0473

Accepted at ICLR 2015 as oral presentation

cs.CL

20140901

20160519

1409.0473

15

RNNenc-30 Figure 2: The BLEU scores of the generated translations on the test set with respect to the lengths of the sentences. The results are on the full test set which includes sentences having unknown words to the models. 2012 and news-test-2013 to make a development (validation) set, and evaluate the models on the test set (news-test-2014) from WMT ’14, which consists of 3003 sentences not present in the training data. After a usual tokenization6, we use a shortlist of 30,000 most frequent words in each language to train our models. Any word not included in the shortlist is mapped to a special token ( [UNK ]). We do not apply any other special preprocessing, such as lowercasing or stemming, to the data. 4.2 M ODELS We train two types of models. The first one is an RNN Encoder–Decoder (RNNencdec, Cho et al. , 2014a), and the other is the proposed model, to which we refer as RNNsearch. We train each model twice: first with the sentences of length up to 30 words (RNNencdec-30, RNNsearch-30) and then

1409.0473

Neural Machine Translation by Jointly Learning to Align and Translate

Neural machine translation is a recently proposed approach to machine translation. Unlike the traditional statistical machine translation, the neural machine translation aims at building a single neural network that can be jointly tuned to maximize the translation performance. The models proposed recently for neural machine translation often belong to a family of encoder-decoders and consists of an encoder that encodes a source sentence into a fixed-length vector from which a decoder generates a translation. In this paper, we conjecture that the use of a fixed-length vector is a bottleneck in improving the performance of this basic encoder-decoder architecture, and propose to extend this by allowing a model to automatically (soft-)search for parts of a source sentence that are relevant to predicting a target word, without having to form these parts as a hard segment explicitly. With this new approach, we achieve a translation performance comparable to the existing state-of-the-art phrase-based system on the task of English-to-French translation. Furthermore, qualitative analysis reveals that the (soft-)alignments found by the model agree well with our intuition.

http://arxiv.org/pdf/1409.0473

Accepted at ICLR 2015 as oral presentation

cs.CL

20140901

20160519

1409.0473

16

twice: first with the sentences of length up to 30 words (RNNencdec-30, RNNsearch-30) and then with the sentences of length up to 50 word (RNNencdec-50, RNNsearch-50). The encoder and decoder of the RNNencdec have 1000 hidden units each.7The encoder of the RNNsearch consists of forward and backward recurrent neural networks (RNN) each having 1000 hidden units. Its decoder has 1000 hidden units. In both cases, we use a multilayer network with a single maxout (Goodfellow et al. , 2013) hidden layer to compute the conditional probability of each target word (Pascanu et al. , 2014). We use a minibatch stochastic gradient descent (SGD) algorithm together with Adadelta (Zeiler, 2012) to train each model. Each SGD update direction is computed using a minibatch of 80 sentences. We trained each model for approximately 5 days. Once a model is trained, we use a beam search to find a translation that approximately maximizes the conditional probability (see, e.g., Graves, 2012; Boulanger-Lewandowski et al. , 2013). Sutskever et al. (2014) used this approach to generate translations from their neural machine translation model.

1409.0473

Neural Machine Translation by Jointly Learning to Align and Translate

Neural machine translation is a recently proposed approach to machine translation. Unlike the traditional statistical machine translation, the neural machine translation aims at building a single neural network that can be jointly tuned to maximize the translation performance. The models proposed recently for neural machine translation often belong to a family of encoder-decoders and consists of an encoder that encodes a source sentence into a fixed-length vector from which a decoder generates a translation. In this paper, we conjecture that the use of a fixed-length vector is a bottleneck in improving the performance of this basic encoder-decoder architecture, and propose to extend this by allowing a model to automatically (soft-)search for parts of a source sentence that are relevant to predicting a target word, without having to form these parts as a hard segment explicitly. With this new approach, we achieve a translation performance comparable to the existing state-of-the-art phrase-based system on the task of English-to-French translation. Furthermore, qualitative analysis reveals that the (soft-)alignments found by the model agree well with our intuition.

http://arxiv.org/pdf/1409.0473

Accepted at ICLR 2015 as oral presentation

cs.CL

20140901

20160519

1409.0473

17

et al. (2014) used this approach to generate translations from their neural machine translation model. For more details on the architectures of the models and training procedure used in the experiments, see Appendices A and B. 5 R ESULTS 5.1 Q UANTITATIVE RESULTS In Table 1, we list the translation performances measured in BLEU score. It is clear from the table that in all the cases, the proposed RNNsearch outperforms the conventional RNNencdec. More importantly, the performance of the RNNsearch is as high as that of the conventional phrase-based translation system (Moses), when only the sentences consisting of known words are considered. This is a significant achievement, considering that Moses uses a separate monolingual corpus (418M words) in addition to the parallel corpora we used to train the RNNsearch and RNNencdec. 6We used the tokenization script from the open-source machine translation package, Moses. 7In this paper, by a ’hidden unit’, we always mean the gated hidden unit (see Appendix A.1.1). 5 Published as a conference paper at ICLR 2015 The agreement on the European Economic Area was signed in August 1992 . <end> L' accord

1409.0473

Neural Machine Translation by Jointly Learning to Align and Translate

Neural machine translation is a recently proposed approach to machine translation. Unlike the traditional statistical machine translation, the neural machine translation aims at building a single neural network that can be jointly tuned to maximize the translation performance. The models proposed recently for neural machine translation often belong to a family of encoder-decoders and consists of an encoder that encodes a source sentence into a fixed-length vector from which a decoder generates a translation. In this paper, we conjecture that the use of a fixed-length vector is a bottleneck in improving the performance of this basic encoder-decoder architecture, and propose to extend this by allowing a model to automatically (soft-)search for parts of a source sentence that are relevant to predicting a target word, without having to form these parts as a hard segment explicitly. With this new approach, we achieve a translation performance comparable to the existing state-of-the-art phrase-based system on the task of English-to-French translation. Furthermore, qualitative analysis reveals that the (soft-)alignments found by the model agree well with our intuition.

http://arxiv.org/pdf/1409.0473

Accepted at ICLR 2015 as oral presentation

cs.CL

20140901

20160519

1409.0473

18

The agreement on the European Economic Area was signed in August 1992 . <end> L' accord sur la zone économique européenne a été signé en août 1992 . <end> It should be noted that the marine environment is the least known of environments . <end> Il convient de noter que l' environnement marin est le moins connu de l' environnement . <end> (a) (b) Destruction of the equipment means that Syria can no longer produce new chemical weapons . <end> La destruction de l' équipement signifie que la Syrie ne peut plus produire de nouvelles armes chimiques . <end> " This will change my future with my family , " the man said . <end> " Cela

1409.0473

Neural Machine Translation by Jointly Learning to Align and Translate

Neural machine translation is a recently proposed approach to machine translation. Unlike the traditional statistical machine translation, the neural machine translation aims at building a single neural network that can be jointly tuned to maximize the translation performance. The models proposed recently for neural machine translation often belong to a family of encoder-decoders and consists of an encoder that encodes a source sentence into a fixed-length vector from which a decoder generates a translation. In this paper, we conjecture that the use of a fixed-length vector is a bottleneck in improving the performance of this basic encoder-decoder architecture, and propose to extend this by allowing a model to automatically (soft-)search for parts of a source sentence that are relevant to predicting a target word, without having to form these parts as a hard segment explicitly. With this new approach, we achieve a translation performance comparable to the existing state-of-the-art phrase-based system on the task of English-to-French translation. Furthermore, qualitative analysis reveals that the (soft-)alignments found by the model agree well with our intuition.

http://arxiv.org/pdf/1409.0473

Accepted at ICLR 2015 as oral presentation

cs.CL

20140901

20160519

1409.0473

19

" This will change my future with my family , " the man said . <end> " Cela va changer mon avenir avec ma famille " , a dit l' homme . <end> (c) (d) Figure 3: Four sample alignments found by RNNsearch-50. The x-axis and y-axis of each plot correspond to the words in the source sentence (English) and the generated translation (French), respectively. Each pixel shows the weight ijof the annotation of the j-th source word for the i-th target word (see Eq. (6)), in grayscale ( 0: black, 1: white). (a) an arbitrary sentence. (b–d) three randomly selected samples among the sentences without any unknown words and of length between 10 and 20 words from the test set. One of the motivations behind the proposed approach was the use of a fixed-length context vector in the basic encoder–decoder approach. We conjectured that this limitation may make the basic

1409.0473

Neural Machine Translation by Jointly Learning to Align and Translate

Neural machine translation is a recently proposed approach to machine translation. Unlike the traditional statistical machine translation, the neural machine translation aims at building a single neural network that can be jointly tuned to maximize the translation performance. The models proposed recently for neural machine translation often belong to a family of encoder-decoders and consists of an encoder that encodes a source sentence into a fixed-length vector from which a decoder generates a translation. In this paper, we conjecture that the use of a fixed-length vector is a bottleneck in improving the performance of this basic encoder-decoder architecture, and propose to extend this by allowing a model to automatically (soft-)search for parts of a source sentence that are relevant to predicting a target word, without having to form these parts as a hard segment explicitly. With this new approach, we achieve a translation performance comparable to the existing state-of-the-art phrase-based system on the task of English-to-French translation. Furthermore, qualitative analysis reveals that the (soft-)alignments found by the model agree well with our intuition.

http://arxiv.org/pdf/1409.0473

Accepted at ICLR 2015 as oral presentation

cs.CL

20140901

20160519

1409.0473

20

in the basic encoder–decoder approach. We conjectured that this limitation may make the basic encoder–decoder approach to underperform with long sentences. In Fig. 2, we see that the performance of RNNencdec dramatically drops as the length of the sentences increases. On the other hand, both RNNsearch-30 and RNNsearch-50 are more robust to the length of the sentences. RNNsearch50, especially, shows no performance deterioration even with sentences of length 50 or more. This superiority of the proposed model over the basic encoder–decoder is further confirmed by the fact that the RNNsearch-30 even outperforms RNNencdec-50 (see Table 1). 6 Published as a conference paper at ICLR 2015 Model All No UNK RNNencdec-30 13.93 24.19 RNNsearch-30 21.50 31.44 RNNencdec-50 17.82 26.71 RNNsearch-50 26.75 34.16 RNNsearch-50?28.45 36.15 Moses 33.30 35.63Table 1: BLEU scores of the trained models computed on the test set. The second and third columns show respectively the scores on all the sentences and,

1409.0473

Neural Machine Translation by Jointly Learning to Align and Translate

Neural machine translation is a recently proposed approach to machine translation. Unlike the traditional statistical machine translation, the neural machine translation aims at building a single neural network that can be jointly tuned to maximize the translation performance. The models proposed recently for neural machine translation often belong to a family of encoder-decoders and consists of an encoder that encodes a source sentence into a fixed-length vector from which a decoder generates a translation. In this paper, we conjecture that the use of a fixed-length vector is a bottleneck in improving the performance of this basic encoder-decoder architecture, and propose to extend this by allowing a model to automatically (soft-)search for parts of a source sentence that are relevant to predicting a target word, without having to form these parts as a hard segment explicitly. With this new approach, we achieve a translation performance comparable to the existing state-of-the-art phrase-based system on the task of English-to-French translation. Furthermore, qualitative analysis reveals that the (soft-)alignments found by the model agree well with our intuition.

http://arxiv.org/pdf/1409.0473

Accepted at ICLR 2015 as oral presentation

cs.CL

20140901

20160519

1409.0473

21

show respectively the scores on all the sentences and, on the sentences without any unknown word in themselves and in the reference translations. Note that RNNsearch-50?was trained much longer until the performance on the development set stopped improving. () We disallowed the models to generate [UNK] tokens when only the sentences having no unknown words were evaluated (last column). 5.2 Q UALITATIVE ANALYSIS 5.2.1 A LIGNMENT The proposed approach provides an intuitive way to inspect the (soft-)alignment between the words in a generated translation and those in a source sentence. This is done by visualizing the annotation weightsijfrom Eq. (6), as in Fig. 3. Each row of a matrix in each plot indicates the weights associated with the annotations. From this we see which positions in the source sentence were considered more important when generating the target word. We can see from the alignments in Fig. 3 that the alignment of words between English and French is largely monotonic. We see strong weights along the diagonal of each matrix. However, we also observe a number of non-trivial, non-monotonic alignments. Adjectives and nouns are typically ordered differently between French and English, and we see an example in Fig. 3 (a). From this

1409.0473

Neural Machine Translation by Jointly Learning to Align and Translate

Neural machine translation is a recently proposed approach to machine translation. Unlike the traditional statistical machine translation, the neural machine translation aims at building a single neural network that can be jointly tuned to maximize the translation performance. The models proposed recently for neural machine translation often belong to a family of encoder-decoders and consists of an encoder that encodes a source sentence into a fixed-length vector from which a decoder generates a translation. In this paper, we conjecture that the use of a fixed-length vector is a bottleneck in improving the performance of this basic encoder-decoder architecture, and propose to extend this by allowing a model to automatically (soft-)search for parts of a source sentence that are relevant to predicting a target word, without having to form these parts as a hard segment explicitly. With this new approach, we achieve a translation performance comparable to the existing state-of-the-art phrase-based system on the task of English-to-French translation. Furthermore, qualitative analysis reveals that the (soft-)alignments found by the model agree well with our intuition.

http://arxiv.org/pdf/1409.0473

Accepted at ICLR 2015 as oral presentation

cs.CL

20140901

20160519

1409.0473

22

ordered differently between French and English, and we see an example in Fig. 3 (a). From this figure, we see that the model correctly translates a phrase [European Economic Area] into [zone ´economique europ ´een]. The RNNsearch was able to correctly align [zone] with [Area], jumping over the two words ([European] and [Economic]), and then looked one word back at a time to complete the whole phrase [zone ´economique europ ´eenne]. The strength of the soft-alignment, opposed to a hard-alignment, is evident, for instance, from Fig. 3 (d). Consider the source phrase [the man] which was translated into [l’ homme]. Any hard alignment will map [the] to [l’] and [man] to [homme]. This is not helpful for translation, as one must consider the word following [the] to determine whether it should be translated into [le], [la], [les] or [l’]. Our soft-alignment solves this issue naturally by letting the model look at both [the] and [man], and in this example, we see that the model was able to correctly translate [the] into [l’]. We

1409.0473

Neural Machine Translation by Jointly Learning to Align and Translate

Neural machine translation is a recently proposed approach to machine translation. Unlike the traditional statistical machine translation, the neural machine translation aims at building a single neural network that can be jointly tuned to maximize the translation performance. The models proposed recently for neural machine translation often belong to a family of encoder-decoders and consists of an encoder that encodes a source sentence into a fixed-length vector from which a decoder generates a translation. In this paper, we conjecture that the use of a fixed-length vector is a bottleneck in improving the performance of this basic encoder-decoder architecture, and propose to extend this by allowing a model to automatically (soft-)search for parts of a source sentence that are relevant to predicting a target word, without having to form these parts as a hard segment explicitly. With this new approach, we achieve a translation performance comparable to the existing state-of-the-art phrase-based system on the task of English-to-French translation. Furthermore, qualitative analysis reveals that the (soft-)alignments found by the model agree well with our intuition.

http://arxiv.org/pdf/1409.0473

Accepted at ICLR 2015 as oral presentation

cs.CL

20140901

20160519

1409.0473

23

[man], and in this example, we see that the model was able to correctly translate [the] into [l’]. We observe similar behaviors in all the presented cases in Fig. 3. An additional benefit of the soft alignment is that it naturally deals with source and target phrases of different lengths, without requiring a counter-intuitive way of mapping some words to or from nowhere ([NULL]) (see, e.g., Chapters 4 and 5 of Koehn, 2010). 5.2.2 L ONG SENTENCES As clearly visible from Fig. 2 the proposed model (RNNsearch) is much better than the conventional model (RNNencdec) at translating long sentences. This is likely due to the fact that the RNNsearch does not require encoding a long sentence into a fixed-length vector perfectly, but only accurately encoding the parts of the input sentence that surround a particular word. As an example, consider this source sentence from the test set: An admitting privilege is the right of a doctor to admit a patient to a hospital or a medical centre tocarry outadiagnosis oraprocedure, based onhisstatus asa health care worker atahospital. The RNNencdec-50 translated this sentence into:

1409.0473

Neural Machine Translation by Jointly Learning to Align and Translate

Neural machine translation is a recently proposed approach to machine translation. Unlike the traditional statistical machine translation, the neural machine translation aims at building a single neural network that can be jointly tuned to maximize the translation performance. The models proposed recently for neural machine translation often belong to a family of encoder-decoders and consists of an encoder that encodes a source sentence into a fixed-length vector from which a decoder generates a translation. In this paper, we conjecture that the use of a fixed-length vector is a bottleneck in improving the performance of this basic encoder-decoder architecture, and propose to extend this by allowing a model to automatically (soft-)search for parts of a source sentence that are relevant to predicting a target word, without having to form these parts as a hard segment explicitly. With this new approach, we achieve a translation performance comparable to the existing state-of-the-art phrase-based system on the task of English-to-French translation. Furthermore, qualitative analysis reveals that the (soft-)alignments found by the model agree well with our intuition.

http://arxiv.org/pdf/1409.0473

Accepted at ICLR 2015 as oral presentation

cs.CL

20140901

20160519

1409.0473

24

health care worker atahospital. The RNNencdec-50 translated this sentence into: Un privil `ege d’admission est le droit d’un m ´edecin de reconna ˆıtre un patient `a l’hˆopital ou un centre m ´edical d’un diagnostic oudeprendre undiagnostic en fonction deson´etatdesant´e. 7 Published as a conference paper at ICLR 2015 The RNNencdec-50 correctly translated the source sentence until [a medical center]. However, from there on (underlined), it deviated from the original meaning of the source sentence. For instance, it replaced [based on his status as a health care worker at a hospital] in the source sentence with [en fonction de son ´etat de sant ´e] (“based on his state of health”). On the other hand, the RNNsearch-50 generated the following correct translation, preserving the whole meaning of the input sentence without omitting any details: Un privil `ege d’admission est le droit d’un m ´edecin d’admettre un patient `a un

1409.0473

Neural Machine Translation by Jointly Learning to Align and Translate

Neural machine translation is a recently proposed approach to machine translation. Unlike the traditional statistical machine translation, the neural machine translation aims at building a single neural network that can be jointly tuned to maximize the translation performance. The models proposed recently for neural machine translation often belong to a family of encoder-decoders and consists of an encoder that encodes a source sentence into a fixed-length vector from which a decoder generates a translation. In this paper, we conjecture that the use of a fixed-length vector is a bottleneck in improving the performance of this basic encoder-decoder architecture, and propose to extend this by allowing a model to automatically (soft-)search for parts of a source sentence that are relevant to predicting a target word, without having to form these parts as a hard segment explicitly. With this new approach, we achieve a translation performance comparable to the existing state-of-the-art phrase-based system on the task of English-to-French translation. Furthermore, qualitative analysis reveals that the (soft-)alignments found by the model agree well with our intuition.

http://arxiv.org/pdf/1409.0473

Accepted at ICLR 2015 as oral presentation

cs.CL

20140901

20160519

1409.0473

25

Un privil `ege d’admission est le droit d’un m ´edecin d’admettre un patient `a un hˆopital ou un centre m ´edical pour effectuer undiagnostic ouuneproc´edure, selon sonstatut detravailleur dessoins desant´e`al’hˆopital. Let us consider another sentence from the test set: This kind of experience is part of Disney’s efforts to ”extend the lifetime of its series and build new relationships with audiences viadigital platforms that are becoming ever more important, ” headded. The translation by the RNNencdec-50 is Ce type d’exp ´erience fait partie des initiatives du Disney pour ”prolonger la dur ´ee de vie de ses nouvelles et de d ´evelopper des liens avec les lecteurs num´eriques qui deviennent plus complexes. As with the previous example, the RNNencdec began deviating from the actual meaning of the source sentence after generating approximately 30 words (see the underlined phrase). After that

1409.0473

Neural Machine Translation by Jointly Learning to Align and Translate

Neural machine translation is a recently proposed approach to machine translation. Unlike the traditional statistical machine translation, the neural machine translation aims at building a single neural network that can be jointly tuned to maximize the translation performance. The models proposed recently for neural machine translation often belong to a family of encoder-decoders and consists of an encoder that encodes a source sentence into a fixed-length vector from which a decoder generates a translation. In this paper, we conjecture that the use of a fixed-length vector is a bottleneck in improving the performance of this basic encoder-decoder architecture, and propose to extend this by allowing a model to automatically (soft-)search for parts of a source sentence that are relevant to predicting a target word, without having to form these parts as a hard segment explicitly. With this new approach, we achieve a translation performance comparable to the existing state-of-the-art phrase-based system on the task of English-to-French translation. Furthermore, qualitative analysis reveals that the (soft-)alignments found by the model agree well with our intuition.

http://arxiv.org/pdf/1409.0473

Accepted at ICLR 2015 as oral presentation

cs.CL

20140901

20160519

1409.0473

26

As with the previous example, the RNNencdec began deviating from the actual meaning of the source sentence after generating approximately 30 words (see the underlined phrase). After that point, the quality of the translation deteriorates, with basic mistakes such as the lack of a closing quotation mark. Again, the RNNsearch-50 was able to translate this long sentence correctly: Ce genre d’exp ´erience fait partie des efforts de Disney pour ”prolonger la dur ´ee de vie de ses s ´eries et cr ´eer de nouvelles relations avec des publics viades plateformes num´eriques deplus enplus importantes”, a-t-il ajout ´e. In conjunction with the quantitative results presented already, these qualitative observations confirm our hypotheses that the RNNsearch architecture enables far more reliable translation of long sentences than the standard RNNencdec model. In Appendix C, we provide a few more sample translations of long source sentences generated by the RNNencdec-50, RNNsearch-50 and Google Translate along with the reference translations. 6 R ELATED WORK 6.1 L EARNING TO ALIGN A similar approach of aligning an output symbol with an input symbol was proposed recently by

1409.0473

Neural Machine Translation by Jointly Learning to Align and Translate

Neural machine translation is a recently proposed approach to machine translation. Unlike the traditional statistical machine translation, the neural machine translation aims at building a single neural network that can be jointly tuned to maximize the translation performance. The models proposed recently for neural machine translation often belong to a family of encoder-decoders and consists of an encoder that encodes a source sentence into a fixed-length vector from which a decoder generates a translation. In this paper, we conjecture that the use of a fixed-length vector is a bottleneck in improving the performance of this basic encoder-decoder architecture, and propose to extend this by allowing a model to automatically (soft-)search for parts of a source sentence that are relevant to predicting a target word, without having to form these parts as a hard segment explicitly. With this new approach, we achieve a translation performance comparable to the existing state-of-the-art phrase-based system on the task of English-to-French translation. Furthermore, qualitative analysis reveals that the (soft-)alignments found by the model agree well with our intuition.

http://arxiv.org/pdf/1409.0473

Accepted at ICLR 2015 as oral presentation

cs.CL

20140901

20160519

1409.0473

27

6 R ELATED WORK 6.1 L EARNING TO ALIGN A similar approach of aligning an output symbol with an input symbol was proposed recently by Graves (2013) in the context of handwriting synthesis. Handwriting synthesis is a task where the model is asked to generate handwriting of a given sequence of characters. In his work, he used a mixture of Gaussian kernels to compute the weights of the annotations, where the location, width and mixture coefficient of each kernel was predicted from an alignment model. More specifically, his alignment was restricted to predict the location such that the location increases monotonically. The main difference from our approach is that, in (Graves, 2013), the modes of the weights of the annotations only move in one direction. In the context of machine translation, this is a severe limitation, as (long-distance) reordering is often needed to generate a grammatically correct translation (for instance, English-to-German). Our approach, on the other hand, requires computing the annotation weight of every word in the source sentence for each word in the translation. This drawback is not severe with the task of translation in which most of input and output sentences are only 15–40 words. However, this may limit the applicability of the proposed scheme to other tasks. 8

1409.0473

Neural Machine Translation by Jointly Learning to Align and Translate

Neural machine translation is a recently proposed approach to machine translation. Unlike the traditional statistical machine translation, the neural machine translation aims at building a single neural network that can be jointly tuned to maximize the translation performance. The models proposed recently for neural machine translation often belong to a family of encoder-decoders and consists of an encoder that encodes a source sentence into a fixed-length vector from which a decoder generates a translation. In this paper, we conjecture that the use of a fixed-length vector is a bottleneck in improving the performance of this basic encoder-decoder architecture, and propose to extend this by allowing a model to automatically (soft-)search for parts of a source sentence that are relevant to predicting a target word, without having to form these parts as a hard segment explicitly. With this new approach, we achieve a translation performance comparable to the existing state-of-the-art phrase-based system on the task of English-to-French translation. Furthermore, qualitative analysis reveals that the (soft-)alignments found by the model agree well with our intuition.

http://arxiv.org/pdf/1409.0473

Accepted at ICLR 2015 as oral presentation

cs.CL

20140901

20160519

1409.0473

28

translation in which most of input and output sentences are only 15–40 words. However, this may limit the applicability of the proposed scheme to other tasks. 8 Published as a conference paper at ICLR 2015 6.2 N EURAL NETWORKS FOR MACHINE TRANSLATION Since Bengio et al. (2003) introduced a neural probabilistic language model which uses a neural network to model the conditional probability of a word given a fixed number of the preceding words, neural networks have widely been used in machine translation. However, the role of neural networks has been largely limited to simply providing a single feature to an existing statistical machine translation system or to re-rank a list of candidate translations provided by an existing system. For instance, Schwenk (2012) proposed using a feedforward neural network to compute the score of a pair of source and target phrases and to use the score as an additional feature in the phrase-based statistical machine translation system. More recently, Kalchbrenner and Blunsom (2013) and Devlin et al. (2014) reported the successful use of the neural networks as a sub-component of the existing translation system. Traditionally, a neural network trained as a target-side language model has been used to rescore or rerank a list of candidate translations (see, e.g., Schwenk et al. , 2006).

1409.0473

Neural Machine Translation by Jointly Learning to Align and Translate

Neural machine translation is a recently proposed approach to machine translation. Unlike the traditional statistical machine translation, the neural machine translation aims at building a single neural network that can be jointly tuned to maximize the translation performance. The models proposed recently for neural machine translation often belong to a family of encoder-decoders and consists of an encoder that encodes a source sentence into a fixed-length vector from which a decoder generates a translation. In this paper, we conjecture that the use of a fixed-length vector is a bottleneck in improving the performance of this basic encoder-decoder architecture, and propose to extend this by allowing a model to automatically (soft-)search for parts of a source sentence that are relevant to predicting a target word, without having to form these parts as a hard segment explicitly. With this new approach, we achieve a translation performance comparable to the existing state-of-the-art phrase-based system on the task of English-to-French translation. Furthermore, qualitative analysis reveals that the (soft-)alignments found by the model agree well with our intuition.

http://arxiv.org/pdf/1409.0473

Accepted at ICLR 2015 as oral presentation

cs.CL

20140901

20160519

1409.0473

29

used to rescore or rerank a list of candidate translations (see, e.g., Schwenk et al. , 2006). Although the above approaches were shown to improve the translation performance over the stateof-the-art machine translation systems, we are more interested in a more ambitious objective of designing a completely new translation system based on neural networks. The neural machine translation approach we consider in this paper is therefore a radical departure from these earlier works. Rather than using a neural network as a part of the existing system, our model works on its own and generates a translation from a source sentence directly. 7 C ONCLUSION The conventional approach to neural machine translation, called an encoder–decoder approach, encodes a whole input sentence into a fixed-length vector from which a translation will be decoded. We conjectured that the use of a fixed-length context vector is problematic for translating long sentences, based on a recent empirical study reported by Cho et al. (2014b) and Pouget-Abadie et al. (2014). In this paper, we proposed a novel architecture that addresses this issue. We extended the basic encoder–decoder by letting a model (soft-)search for a set of input words, or their annotations computed by an encoder, when generating each target word. This frees the model from having to encode

1409.0473

Neural Machine Translation by Jointly Learning to Align and Translate

Neural machine translation is a recently proposed approach to machine translation. Unlike the traditional statistical machine translation, the neural machine translation aims at building a single neural network that can be jointly tuned to maximize the translation performance. The models proposed recently for neural machine translation often belong to a family of encoder-decoders and consists of an encoder that encodes a source sentence into a fixed-length vector from which a decoder generates a translation. In this paper, we conjecture that the use of a fixed-length vector is a bottleneck in improving the performance of this basic encoder-decoder architecture, and propose to extend this by allowing a model to automatically (soft-)search for parts of a source sentence that are relevant to predicting a target word, without having to form these parts as a hard segment explicitly. With this new approach, we achieve a translation performance comparable to the existing state-of-the-art phrase-based system on the task of English-to-French translation. Furthermore, qualitative analysis reveals that the (soft-)alignments found by the model agree well with our intuition.

http://arxiv.org/pdf/1409.0473

Accepted at ICLR 2015 as oral presentation

cs.CL

20140901

20160519

1409.0473

30

a whole source sentence into a fixed-length vector, and also lets the model focus only on information relevant to the generation of the next target word. This has a major positive impact on the ability of the neural machine translation system to yield good results on longer sentences. Unlike with the traditional machine translation systems, all of the pieces of the translation system, including the alignment mechanism, are jointly trained towards a better log-probability of producing correct translations. We tested the proposed model, called RNNsearch, on the task of English-to-French translation. The experiment revealed that the proposed RNNsearch outperforms the conventional encoder–decoder model (RNNencdec) significantly, regardless of the sentence length and that it is much more robust to the length of a source sentence. From the qualitative analysis where we investigated the (soft-)alignment generated by the RNNsearch, we were able to conclude that the model can correctly align each target word with the relevant words, or their annotations, in the source sentence as it generated a correct translation. Perhaps more importantly, the proposed approach achieved a translation performance comparable to the existing phrase-based statistical machine translation. It is a striking result, considering that the proposed architecture, or the whole family of neural machine translation, has only been proposed

1409.0473

Neural Machine Translation by Jointly Learning to Align and Translate

Neural machine translation is a recently proposed approach to machine translation. Unlike the traditional statistical machine translation, the neural machine translation aims at building a single neural network that can be jointly tuned to maximize the translation performance. The models proposed recently for neural machine translation often belong to a family of encoder-decoders and consists of an encoder that encodes a source sentence into a fixed-length vector from which a decoder generates a translation. In this paper, we conjecture that the use of a fixed-length vector is a bottleneck in improving the performance of this basic encoder-decoder architecture, and propose to extend this by allowing a model to automatically (soft-)search for parts of a source sentence that are relevant to predicting a target word, without having to form these parts as a hard segment explicitly. With this new approach, we achieve a translation performance comparable to the existing state-of-the-art phrase-based system on the task of English-to-French translation. Furthermore, qualitative analysis reveals that the (soft-)alignments found by the model agree well with our intuition.

http://arxiv.org/pdf/1409.0473

Accepted at ICLR 2015 as oral presentation

cs.CL

20140901

20160519

1409.0473

31

the existing phrase-based statistical machine translation. It is a striking result, considering that the proposed architecture, or the whole family of neural machine translation, has only been proposed as recently as this year. We believe the architecture proposed here is a promising step toward better machine translation and a better understanding of natural languages in general. One of challenges left for the future is to better handle unknown, or rare words. This will be required for the model to be more widely used and to match the performance of current state-of-the-art machine translation systems in all contexts. 9 Published as a conference paper at ICLR 2015 ACKNOWLEDGMENTS The authors would like to thank the developers of Theano (Bergstra et al. , 2010; Bastien et al. , 2012). We acknowledge the support of the following agencies for research funding and computing support: NSERC, Calcul Qu ´ebec, Compute Canada, the Canada Research Chairs and CIFAR. Bahdanau thanks the support from Planet Intelligent Systems GmbH. We also thank Felix Hill, Bart van Merri ´enboer, Jean Pouget-Abadie, Coline Devin and Tae-Ho Kim. REFERENCES Axelrod, A., He, X., and Gao, J. (2011). Domain adaptation via pseudo in-domain data selection.

1409.0473

Neural Machine Translation by Jointly Learning to Align and Translate

Neural machine translation is a recently proposed approach to machine translation. Unlike the traditional statistical machine translation, the neural machine translation aims at building a single neural network that can be jointly tuned to maximize the translation performance. The models proposed recently for neural machine translation often belong to a family of encoder-decoders and consists of an encoder that encodes a source sentence into a fixed-length vector from which a decoder generates a translation. In this paper, we conjecture that the use of a fixed-length vector is a bottleneck in improving the performance of this basic encoder-decoder architecture, and propose to extend this by allowing a model to automatically (soft-)search for parts of a source sentence that are relevant to predicting a target word, without having to form these parts as a hard segment explicitly. With this new approach, we achieve a translation performance comparable to the existing state-of-the-art phrase-based system on the task of English-to-French translation. Furthermore, qualitative analysis reveals that the (soft-)alignments found by the model agree well with our intuition.

http://arxiv.org/pdf/1409.0473

Accepted at ICLR 2015 as oral presentation

cs.CL

20140901

20160519

1409.0473

32

REFERENCES Axelrod, A., He, X., and Gao, J. (2011). Domain adaptation via pseudo in-domain data selection. InProceedings of the ACL Conference on Empirical Methods in Natural Language Processing (EMNLP) , pages 355–362. Association for Computational Linguistics. Bastien, F., Lamblin, P., Pascanu, R., Bergstra, J., Goodfellow, I. J., Bergeron, A., Bouchard, N., and Bengio, Y . (2012). Theano: new features and speed improvements. Deep Learning and Unsupervised Feature Learning NIPS 2012 Workshop. Bengio, Y ., Simard, P., and Frasconi, P. (1994). Learning long-term dependencies with gradient descent is difficult. IEEE Transactions on Neural Networks ,5(2), 157–166. Bengio, Y ., Ducharme, R., Vincent, P., and Janvin, C. (2003). A neural probabilistic language model. J. Mach. Learn. Res. ,3, 1137–1155.

1409.0473

Neural Machine Translation by Jointly Learning to Align and Translate

Neural machine translation is a recently proposed approach to machine translation. Unlike the traditional statistical machine translation, the neural machine translation aims at building a single neural network that can be jointly tuned to maximize the translation performance. The models proposed recently for neural machine translation often belong to a family of encoder-decoders and consists of an encoder that encodes a source sentence into a fixed-length vector from which a decoder generates a translation. In this paper, we conjecture that the use of a fixed-length vector is a bottleneck in improving the performance of this basic encoder-decoder architecture, and propose to extend this by allowing a model to automatically (soft-)search for parts of a source sentence that are relevant to predicting a target word, without having to form these parts as a hard segment explicitly. With this new approach, we achieve a translation performance comparable to the existing state-of-the-art phrase-based system on the task of English-to-French translation. Furthermore, qualitative analysis reveals that the (soft-)alignments found by the model agree well with our intuition.

http://arxiv.org/pdf/1409.0473

Accepted at ICLR 2015 as oral presentation

cs.CL

20140901

20160519

1409.0473

33

J. Mach. Learn. Res. ,3, 1137–1155. Bergstra, J., Breuleux, O., Bastien, F., Lamblin, P., Pascanu, R., Desjardins, G., Turian, J., WardeFarley, D., and Bengio, Y . (2010). Theano: a CPU and GPU math expression compiler. In Proceedings of the Python for Scientific Computing Conference (SciPy) . Oral Presentation. Boulanger-Lewandowski, N., Bengio, Y ., and Vincent, P. (2013). Audio chord recognition with recurrent neural networks. In ISMIR . Cho, K., van Merrienboer, B., Gulcehre, C., Bougares, F., Schwenk, H., and Bengio, Y . (2014a). Learning phrase representations using RNN encoder-decoder for statistical machine translation. InProceedings of the Empiricial Methods in Natural Language Processing (EMNLP 2014) . to appear. Cho, K., van Merri ¨enboer, B., Bahdanau, D., and Bengio, Y . (2014b). On the properties of neural machine translation: Encoder–Decoder approaches. In Eighth Workshop on Syntax, Semantics

1409.0473

Neural Machine Translation by Jointly Learning to Align and Translate

Neural machine translation is a recently proposed approach to machine translation. Unlike the traditional statistical machine translation, the neural machine translation aims at building a single neural network that can be jointly tuned to maximize the translation performance. The models proposed recently for neural machine translation often belong to a family of encoder-decoders and consists of an encoder that encodes a source sentence into a fixed-length vector from which a decoder generates a translation. In this paper, we conjecture that the use of a fixed-length vector is a bottleneck in improving the performance of this basic encoder-decoder architecture, and propose to extend this by allowing a model to automatically (soft-)search for parts of a source sentence that are relevant to predicting a target word, without having to form these parts as a hard segment explicitly. With this new approach, we achieve a translation performance comparable to the existing state-of-the-art phrase-based system on the task of English-to-French translation. Furthermore, qualitative analysis reveals that the (soft-)alignments found by the model agree well with our intuition.

http://arxiv.org/pdf/1409.0473

Accepted at ICLR 2015 as oral presentation

cs.CL

20140901

20160519

1409.0473

34

machine translation: Encoder–Decoder approaches. In Eighth Workshop on Syntax, Semantics and Structure in Statistical Translation . to appear. Devlin, J., Zbib, R., Huang, Z., Lamar, T., Schwartz, R., and Makhoul, J. (2014). Fast and robust neural network joint models for statistical machine translation. In Association for Computational Linguistics . Forcada, M. L. and ˜Neco, R. P. (1997). Recursive hetero-associative memories for translation. In J. Mira, R. Moreno-D ´ıaz, and J. Cabestany, editors, Biological and Artificial Computation: From Neuroscience to Technology , volume 1240 of Lecture Notes in Computer Science , pages 453–462. Springer Berlin Heidelberg. Goodfellow, I., Warde-Farley, D., Mirza, M., Courville, A., and Bengio, Y . (2013). Maxout networks. In Proceedings of The 30th International Conference on Machine Learning , pages 1319– 1327. Graves, A. (2012). Sequence transduction with recurrent neural networks. In Proceedings of the 29th International Conference on Machine Learning (ICML 2012) .

1409.0473

Neural Machine Translation by Jointly Learning to Align and Translate

Neural machine translation is a recently proposed approach to machine translation. Unlike the traditional statistical machine translation, the neural machine translation aims at building a single neural network that can be jointly tuned to maximize the translation performance. The models proposed recently for neural machine translation often belong to a family of encoder-decoders and consists of an encoder that encodes a source sentence into a fixed-length vector from which a decoder generates a translation. In this paper, we conjecture that the use of a fixed-length vector is a bottleneck in improving the performance of this basic encoder-decoder architecture, and propose to extend this by allowing a model to automatically (soft-)search for parts of a source sentence that are relevant to predicting a target word, without having to form these parts as a hard segment explicitly. With this new approach, we achieve a translation performance comparable to the existing state-of-the-art phrase-based system on the task of English-to-French translation. Furthermore, qualitative analysis reveals that the (soft-)alignments found by the model agree well with our intuition.

http://arxiv.org/pdf/1409.0473

Accepted at ICLR 2015 as oral presentation

cs.CL

20140901

20160519

1409.0473

35

1327. Graves, A. (2012). Sequence transduction with recurrent neural networks. In Proceedings of the 29th International Conference on Machine Learning (ICML 2012) . Graves, A. (2013). Generating sequences with recurrent neural networks. arXiv: 1308.0850 [cs.NE] . Graves, A., Jaitly, N., and Mohamed, A.-R. (2013). Hybrid speech recognition with deep bidirectional LSTM. In Automatic Speech Recognition and Understanding (ASRU), 2013 IEEE Workshop on , pages 273–278. 10 Published as a conference paper at ICLR 2015 Hermann, K. and Blunsom, P. (2014). Multilingual distributed representations without word alignment. In Proceedings of the Second International Conference on Learning Representations (ICLR 2014) . Hochreiter, S. (1991). Untersuchungen zu dynamischen neuronalen Netzen. Diploma thesis, Institut f¨ur Informatik, Lehrstuhl Prof. Brauer, Technische Universit ¨at M ¨unchen. Hochreiter, S. and Schmidhuber, J. (1997). Long short-term memory. Neural Computation ,9(8), 1735–1780.

1409.0473

Neural Machine Translation by Jointly Learning to Align and Translate

Neural machine translation is a recently proposed approach to machine translation. Unlike the traditional statistical machine translation, the neural machine translation aims at building a single neural network that can be jointly tuned to maximize the translation performance. The models proposed recently for neural machine translation often belong to a family of encoder-decoders and consists of an encoder that encodes a source sentence into a fixed-length vector from which a decoder generates a translation. In this paper, we conjecture that the use of a fixed-length vector is a bottleneck in improving the performance of this basic encoder-decoder architecture, and propose to extend this by allowing a model to automatically (soft-)search for parts of a source sentence that are relevant to predicting a target word, without having to form these parts as a hard segment explicitly. With this new approach, we achieve a translation performance comparable to the existing state-of-the-art phrase-based system on the task of English-to-French translation. Furthermore, qualitative analysis reveals that the (soft-)alignments found by the model agree well with our intuition.

http://arxiv.org/pdf/1409.0473

Accepted at ICLR 2015 as oral presentation

cs.CL

20140901

20160519

1409.0473

36

Hochreiter, S. and Schmidhuber, J. (1997). Long short-term memory. Neural Computation ,9(8), 1735–1780. Kalchbrenner, N. and Blunsom, P. (2013). Recurrent continuous translation models. In Proceedings of the ACL Conference on Empirical Methods in Natural Language Processing (EMNLP) , pages 1700–1709. Association for Computational Linguistics. Koehn, P. (2010). Statistical Machine Translation . Cambridge University Press, New York, NY , USA. Koehn, P., Och, F. J., and Marcu, D. (2003). Statistical phrase-based translation. In Proceedings of the 2003 Conference of the North American Chapter of the Association for Computational Linguistics on Human Language Technology - Volume 1 , NAACL ’03, pages 48–54, Stroudsburg, PA, USA. Association for Computational Linguistics. Pascanu, R., Mikolov, T., and Bengio, Y . (2013a). On the difficulty of training recurrent neural networks. In ICML’2013 . Pascanu, R., Mikolov, T., and Bengio, Y . (2013b). On the difficulty of training recurrent neural

1409.0473

Neural Machine Translation by Jointly Learning to Align and Translate

Neural machine translation is a recently proposed approach to machine translation. Unlike the traditional statistical machine translation, the neural machine translation aims at building a single neural network that can be jointly tuned to maximize the translation performance. The models proposed recently for neural machine translation often belong to a family of encoder-decoders and consists of an encoder that encodes a source sentence into a fixed-length vector from which a decoder generates a translation. In this paper, we conjecture that the use of a fixed-length vector is a bottleneck in improving the performance of this basic encoder-decoder architecture, and propose to extend this by allowing a model to automatically (soft-)search for parts of a source sentence that are relevant to predicting a target word, without having to form these parts as a hard segment explicitly. With this new approach, we achieve a translation performance comparable to the existing state-of-the-art phrase-based system on the task of English-to-French translation. Furthermore, qualitative analysis reveals that the (soft-)alignments found by the model agree well with our intuition.

http://arxiv.org/pdf/1409.0473

Accepted at ICLR 2015 as oral presentation

cs.CL

20140901

20160519

1409.0473

37

Pascanu, R., Mikolov, T., and Bengio, Y . (2013b). On the difficulty of training recurrent neural networks. In Proceedings of the 30th International Conference on Machine Learning (ICML 2013) . Pascanu, R., Gulcehre, C., Cho, K., and Bengio, Y . (2014). How to construct deep recurrent neural networks. In Proceedings of the Second International Conference on Learning Representations (ICLR 2014) . Pouget-Abadie, J., Bahdanau, D., van Merri ¨enboer, B., Cho, K., and Bengio, Y . (2014). Overcoming the curse of sentence length for neural machine translation using automatic segmentation. In Eighth Workshop on Syntax, Semantics and Structure in Statistical Translation . to appear. Schuster, M. and Paliwal, K. K. (1997). Bidirectional recurrent neural networks. Signal Processing, IEEE Transactions on ,45(11), 2673–2681. Schwenk, H. (2012). Continuous space translation models for phrase-based statistical machine translation. In M. Kay and C. Boitet, editors, Proceedings of the 24th International Conference on

1409.0473

Neural Machine Translation by Jointly Learning to Align and Translate

Neural machine translation is a recently proposed approach to machine translation. Unlike the traditional statistical machine translation, the neural machine translation aims at building a single neural network that can be jointly tuned to maximize the translation performance. The models proposed recently for neural machine translation often belong to a family of encoder-decoders and consists of an encoder that encodes a source sentence into a fixed-length vector from which a decoder generates a translation. In this paper, we conjecture that the use of a fixed-length vector is a bottleneck in improving the performance of this basic encoder-decoder architecture, and propose to extend this by allowing a model to automatically (soft-)search for parts of a source sentence that are relevant to predicting a target word, without having to form these parts as a hard segment explicitly. With this new approach, we achieve a translation performance comparable to the existing state-of-the-art phrase-based system on the task of English-to-French translation. Furthermore, qualitative analysis reveals that the (soft-)alignments found by the model agree well with our intuition.

http://arxiv.org/pdf/1409.0473

Accepted at ICLR 2015 as oral presentation

cs.CL

20140901

20160519

1409.0473

38

translation. In M. Kay and C. Boitet, editors, Proceedings of the 24th International Conference on Computational Linguistics (COLIN) , pages 1071–1080. Indian Institute of Technology Bombay. Schwenk, H., Dchelotte, D., and Gauvain, J.-L. (2006). Continuous space language models for statistical machine translation. In Proceedings of the COLING/ACL on Main conference poster sessions , pages 723–730. Association for Computational Linguistics. Sutskever, I., Vinyals, O., and Le, Q. (2014). Sequence to sequence learning with neural networks. InAdvances in Neural Information Processing Systems (NIPS 2014) . Zeiler, M. D. (2012). ADADELTA: An adaptive learning rate method. arXiv: 1212.5701 [cs.LG] . 11 Published as a conference paper at ICLR 2015 A M ODEL ARCHITECTURE A.1 A RCHITECTURAL CHOICES The proposed scheme in Section 3 is a general framework where one can freely define, for instance, the activation functions fof recurrent neural networks (RNN) and the alignment model a. Here, we describe the choices we made for the experiments in this paper.

1409.0473

Neural Machine Translation by Jointly Learning to Align and Translate

Neural machine translation is a recently proposed approach to machine translation. Unlike the traditional statistical machine translation, the neural machine translation aims at building a single neural network that can be jointly tuned to maximize the translation performance. The models proposed recently for neural machine translation often belong to a family of encoder-decoders and consists of an encoder that encodes a source sentence into a fixed-length vector from which a decoder generates a translation. In this paper, we conjecture that the use of a fixed-length vector is a bottleneck in improving the performance of this basic encoder-decoder architecture, and propose to extend this by allowing a model to automatically (soft-)search for parts of a source sentence that are relevant to predicting a target word, without having to form these parts as a hard segment explicitly. With this new approach, we achieve a translation performance comparable to the existing state-of-the-art phrase-based system on the task of English-to-French translation. Furthermore, qualitative analysis reveals that the (soft-)alignments found by the model agree well with our intuition.

http://arxiv.org/pdf/1409.0473

Accepted at ICLR 2015 as oral presentation

cs.CL

20140901

20160519

1409.0473

39

the activation functions fof recurrent neural networks (RNN) and the alignment model a. Here, we describe the choices we made for the experiments in this paper. A.1.1 R ECURRENT NEURAL NETWORK For the activation function fof an RNN, we use the gated hidden unit recently proposed by Cho et al. (2014a). The gated hidden unit is an alternative to the conventional simple units such as an element-wise tanh . This gated unit is similar to a long short-term memory (LSTM) unit proposed earlier by Hochreiter and Schmidhuber (1997), sharing with it the ability to better model and learn long-term dependencies. This is made possible by having computation paths in the unfolded RNN for which the product of derivatives is close to 1. These paths allow gradients to flow backward easily without suffering too much from the vanishing effect (Hochreiter, 1991; Bengio et al. , 1994; Pascanu et al. , 2013a). It is therefore possible to use LSTM units instead of the gated hidden unit described here, as was done in a similar context by Sutskever et al. (2014). The new state siof the RNN employing ngated hidden units8is computed by

1409.0473

Neural Machine Translation by Jointly Learning to Align and Translate

Neural machine translation is a recently proposed approach to machine translation. Unlike the traditional statistical machine translation, the neural machine translation aims at building a single neural network that can be jointly tuned to maximize the translation performance. The models proposed recently for neural machine translation often belong to a family of encoder-decoders and consists of an encoder that encodes a source sentence into a fixed-length vector from which a decoder generates a translation. In this paper, we conjecture that the use of a fixed-length vector is a bottleneck in improving the performance of this basic encoder-decoder architecture, and propose to extend this by allowing a model to automatically (soft-)search for parts of a source sentence that are relevant to predicting a target word, without having to form these parts as a hard segment explicitly. With this new approach, we achieve a translation performance comparable to the existing state-of-the-art phrase-based system on the task of English-to-French translation. Furthermore, qualitative analysis reveals that the (soft-)alignments found by the model agree well with our intuition.

http://arxiv.org/pdf/1409.0473

Accepted at ICLR 2015 as oral presentation

cs.CL

20140901

20160519

1409.0473

40

described here, as was done in a similar context by Sutskever et al. (2014). The new state siof the RNN employing ngated hidden units8is computed by si=f(si1;yi1;ci) = (1zi)si1+zi~si; whereis an element-wise multiplication, and ziis the output of the update gates (see below). The proposed updated state ~siis computed by ~si= tanh (We(yi1) +U[risi1] +Cci); wheree(yi1)2Rmis anm-dimensional embedding of a word yi1, andriis the output of the reset gates (see below). When yiis represented as a 1-of-Kvector,e(yi)is simply a column of an embedding matrix E2RmK. Whenever possible, we omit bias terms to make the equations less cluttered. The update gates ziallow each hidden unit to maintain its previous activation, and the reset gates ri control how much and what information from the previous state should be reset. We compute them by zi=(Wze(yi1) +Uzsi1+Czci);

1409.0473

Neural Machine Translation by Jointly Learning to Align and Translate

Neural machine translation is a recently proposed approach to machine translation. Unlike the traditional statistical machine translation, the neural machine translation aims at building a single neural network that can be jointly tuned to maximize the translation performance. The models proposed recently for neural machine translation often belong to a family of encoder-decoders and consists of an encoder that encodes a source sentence into a fixed-length vector from which a decoder generates a translation. In this paper, we conjecture that the use of a fixed-length vector is a bottleneck in improving the performance of this basic encoder-decoder architecture, and propose to extend this by allowing a model to automatically (soft-)search for parts of a source sentence that are relevant to predicting a target word, without having to form these parts as a hard segment explicitly. With this new approach, we achieve a translation performance comparable to the existing state-of-the-art phrase-based system on the task of English-to-French translation. Furthermore, qualitative analysis reveals that the (soft-)alignments found by the model agree well with our intuition.

http://arxiv.org/pdf/1409.0473

Accepted at ICLR 2015 as oral presentation

cs.CL

20140901

20160519

1409.0473

41

by zi=(Wze(yi1) +Uzsi1+Czci); ri=(Wre(yi1) +Ursi1+Crci); where(
)is a logistic sigmoid function. At each step of the decoder, we compute the output probability (Eq. (4)) as a multi-layered function (Pascanu et al. , 2014). We use a single hidden layer of maxout units (Goodfellow et al. , 2013) and normalize the output probabilities (one for each word) with a softmax function (see Eq. (6)). A.1.2 A LIGNMENT MODEL The alignment model should be designed considering that the model needs to be evaluated TxTy times for each sentence pair of lengths TxandTy. In order to reduce computation, we use a singlelayer multilayer perceptron such that a(si1;hj) =v> atanh (Wasi1+Uahj); whereWa2Rnn;Ua2Rn2nandva2Rnare the weight matrices. Since Uahjdoes not depend oni, we can pre-compute it in advance to minimize the computational cost.

1409.0473

Neural Machine Translation by Jointly Learning to Align and Translate

Neural machine translation is a recently proposed approach to machine translation. Unlike the traditional statistical machine translation, the neural machine translation aims at building a single neural network that can be jointly tuned to maximize the translation performance. The models proposed recently for neural machine translation often belong to a family of encoder-decoders and consists of an encoder that encodes a source sentence into a fixed-length vector from which a decoder generates a translation. In this paper, we conjecture that the use of a fixed-length vector is a bottleneck in improving the performance of this basic encoder-decoder architecture, and propose to extend this by allowing a model to automatically (soft-)search for parts of a source sentence that are relevant to predicting a target word, without having to form these parts as a hard segment explicitly. With this new approach, we achieve a translation performance comparable to the existing state-of-the-art phrase-based system on the task of English-to-French translation. Furthermore, qualitative analysis reveals that the (soft-)alignments found by the model agree well with our intuition.

http://arxiv.org/pdf/1409.0473

Accepted at ICLR 2015 as oral presentation

cs.CL

20140901

20160519

1409.0473

42

depend oni, we can pre-compute it in advance to minimize the computational cost. 8Here, we show the formula of the decoder. The same formula can be used in the encoder by simply ignoring the context vector ciand the related terms. 12 Published as a conference paper at ICLR 2015 A.2 D ETAILED DESCRIPTION OF THE MODEL A.2.1 E NCODER In this section, we describe in detail the architecture of the proposed model (RNNsearch) used in the experiments (see Sec. 4–5). From here on, we omit all bias terms in order to increase readability. The model takes a source sentence of 1-of-K coded word vectors as input x= (x1;:::;x Tx); xi2RKx and outputs a translated sentence of 1-of-K coded word vectors y= (y1;:::;y Ty); yi2RKy; whereKxandKyare the vocabulary sizes of source and target languages, respectively. TxandTy respectively denote the lengths of source and target sentences. First, the forward states of the bidirectional recurrent neural network (BiRNN) are computed: !hi=(

1409.0473

Neural Machine Translation by Jointly Learning to Align and Translate

Neural machine translation is a recently proposed approach to machine translation. Unlike the traditional statistical machine translation, the neural machine translation aims at building a single neural network that can be jointly tuned to maximize the translation performance. The models proposed recently for neural machine translation often belong to a family of encoder-decoders and consists of an encoder that encodes a source sentence into a fixed-length vector from which a decoder generates a translation. In this paper, we conjecture that the use of a fixed-length vector is a bottleneck in improving the performance of this basic encoder-decoder architecture, and propose to extend this by allowing a model to automatically (soft-)search for parts of a source sentence that are relevant to predicting a target word, without having to form these parts as a hard segment explicitly. With this new approach, we achieve a translation performance comparable to the existing state-of-the-art phrase-based system on the task of English-to-French translation. Furthermore, qualitative analysis reveals that the (soft-)alignments found by the model agree well with our intuition.

http://arxiv.org/pdf/1409.0473

Accepted at ICLR 2015 as oral presentation

cs.CL

20140901

20160519

1409.0473

43

respectively denote the lengths of source and target sentences. First, the forward states of the bidirectional recurrent neural network (BiRNN) are computed: !hi=( (1 !zi) !hi1+ !zi !hi, ifi>0 0 , ifi= 0 where !hi= tanh !WExi+ !Uh !ri !hi1i !zi= !WzExi+ !Uz !hi1 !ri= !WrExi+ !Ur !hi1 : E2RmKxis the word embedding matrix. !W; !Wz; !Wr2Rnm, !U; !Uz; !Ur2Rnnare weight matrices. mandnare the word embedding dimensionality and the number of hidden units, respectively. (
)is as usual a logistic sigmoid function. The backward states ( h1;


; hTx)are computed similarly. We share the word embedding matrix Ebetween the forward and backward RNNs, unlike the weight matrices.

1409.0473

Neural Machine Translation by Jointly Learning to Align and Translate

Neural machine translation is a recently proposed approach to machine translation. Unlike the traditional statistical machine translation, the neural machine translation aims at building a single neural network that can be jointly tuned to maximize the translation performance. The models proposed recently for neural machine translation often belong to a family of encoder-decoders and consists of an encoder that encodes a source sentence into a fixed-length vector from which a decoder generates a translation. In this paper, we conjecture that the use of a fixed-length vector is a bottleneck in improving the performance of this basic encoder-decoder architecture, and propose to extend this by allowing a model to automatically (soft-)search for parts of a source sentence that are relevant to predicting a target word, without having to form these parts as a hard segment explicitly. With this new approach, we achieve a translation performance comparable to the existing state-of-the-art phrase-based system on the task of English-to-French translation. Furthermore, qualitative analysis reveals that the (soft-)alignments found by the model agree well with our intuition.

http://arxiv.org/pdf/1409.0473

Accepted at ICLR 2015 as oral presentation

cs.CL

20140901

20160519

1409.0473

44

Ebetween the forward and backward RNNs, unlike the weight matrices. We concatenate the forward and backward states to to obtain the annotations (h1;h2;


;hTx), where hi=" !hi hi# (7) A.2.2 D ECODER The hidden state siof the decoder given the annotations from the encoder is computed by si=(1zi)si1+zi~si; where ~si= tanh (WEy i1+U[risi1] +Cci) zi=(WzEyi1+Uzsi1+Czci) ri=(WrEyi1+Ursi1+Crci) Eis the word embedding matrix for the target language. W;W z;Wr2Rnm,U;Uz;Ur2Rnn, andC;C z;Cr2Rn2nare weights. Again, mandnare the word embedding dimensionality and the number of hidden units, respectively. The initial hidden state s0is computed by s0= tanh Ws h1 ;whereWs2Rnn.

1409.0473

Neural Machine Translation by Jointly Learning to Align and Translate

Neural machine translation is a recently proposed approach to machine translation. Unlike the traditional statistical machine translation, the neural machine translation aims at building a single neural network that can be jointly tuned to maximize the translation performance. The models proposed recently for neural machine translation often belong to a family of encoder-decoders and consists of an encoder that encodes a source sentence into a fixed-length vector from which a decoder generates a translation. In this paper, we conjecture that the use of a fixed-length vector is a bottleneck in improving the performance of this basic encoder-decoder architecture, and propose to extend this by allowing a model to automatically (soft-)search for parts of a source sentence that are relevant to predicting a target word, without having to form these parts as a hard segment explicitly. With this new approach, we achieve a translation performance comparable to the existing state-of-the-art phrase-based system on the task of English-to-French translation. Furthermore, qualitative analysis reveals that the (soft-)alignments found by the model agree well with our intuition.

http://arxiv.org/pdf/1409.0473

Accepted at ICLR 2015 as oral presentation

cs.CL

20140901

20160519

1409.0473

45

tanh Ws h1 ;whereWs2Rnn. The context vector ciare recomputed at each step by the alignment model: ci=TxX j=1ijhj; 13 Published as a conference paper at ICLR 2015 Model Updates (105)Epochs Hours GPU Train NLL Dev. NLL RNNenc-30 8.46 6.4 109 TITAN BLACK 28.1 53.0 RNNenc-50 6.00 4.5 108 Quadro K-6000 44.0 43.6 RNNsearch-30 4.71 3.6 113 TITAN BLACK 26.7 47.2 RNNsearch-50 2.88 2.2 111 Quadro K-6000 40.7 38.1 RNNsearch-50?6.67 5.0 252 Quadro K-6000 36.7 35.2 Table 2: Learning statistics and relevant information. Each update corresponds to updating the parameters once using a single minibatch. One epoch is one pass through the training set. NLL is the average conditional log-probabilities of the sentences in either the training set or the development set. Note that the lengths of the sentences differ. where ij=exp (eij)PTx k=1exp (eik) eij=v>

1409.0473

Neural Machine Translation by Jointly Learning to Align and Translate

Neural machine translation is a recently proposed approach to machine translation. Unlike the traditional statistical machine translation, the neural machine translation aims at building a single neural network that can be jointly tuned to maximize the translation performance. The models proposed recently for neural machine translation often belong to a family of encoder-decoders and consists of an encoder that encodes a source sentence into a fixed-length vector from which a decoder generates a translation. In this paper, we conjecture that the use of a fixed-length vector is a bottleneck in improving the performance of this basic encoder-decoder architecture, and propose to extend this by allowing a model to automatically (soft-)search for parts of a source sentence that are relevant to predicting a target word, without having to form these parts as a hard segment explicitly. With this new approach, we achieve a translation performance comparable to the existing state-of-the-art phrase-based system on the task of English-to-French translation. Furthermore, qualitative analysis reveals that the (soft-)alignments found by the model agree well with our intuition.

http://arxiv.org/pdf/1409.0473

Accepted at ICLR 2015 as oral presentation

cs.CL

20140901

20160519

1409.0473

46

set. Note that the lengths of the sentences differ. where ij=exp (eij)PTx k=1exp (eik) eij=v> atanh (Wasi1+Uahj); andhjis thej-th annotation in the source sentence (see Eq. (7)). va2Rn0;Wa2Rn0nand Ua2Rn02nare weight matrices. Note that the model becomes RNN Encoder–Decoder (Cho et al. , 2014a), if we fix cito !hTx. With the decoder state si1, the context ciand the last generated word yi1, we define the probability of a target word yias p(yijsi;yi1;ci)/exp y> iWoti
; where ti= max~ti;2j1;~ti;2j > j=1;:::;l and~ti;kis thek-th element of a vector ~tiwhich is computed by ~ti=Uosi1+VoEyi1+Coci:

1409.0473

Neural Machine Translation by Jointly Learning to Align and Translate

Neural machine translation is a recently proposed approach to machine translation. Unlike the traditional statistical machine translation, the neural machine translation aims at building a single neural network that can be jointly tuned to maximize the translation performance. The models proposed recently for neural machine translation often belong to a family of encoder-decoders and consists of an encoder that encodes a source sentence into a fixed-length vector from which a decoder generates a translation. In this paper, we conjecture that the use of a fixed-length vector is a bottleneck in improving the performance of this basic encoder-decoder architecture, and propose to extend this by allowing a model to automatically (soft-)search for parts of a source sentence that are relevant to predicting a target word, without having to form these parts as a hard segment explicitly. With this new approach, we achieve a translation performance comparable to the existing state-of-the-art phrase-based system on the task of English-to-French translation. Furthermore, qualitative analysis reveals that the (soft-)alignments found by the model agree well with our intuition.

http://arxiv.org/pdf/1409.0473

Accepted at ICLR 2015 as oral presentation

cs.CL

20140901

20160519

1409.0473

47

and~ti;kis thek-th element of a vector ~tiwhich is computed by ~ti=Uosi1+VoEyi1+Coci: Wo2RKyl,Uo2R2ln,Vo2R2lmandCo2R2l2nare weight matrices. This can be understood as having a deep output (Pascanu et al. , 2014) with a single maxout hidden layer (Goodfellow et al. , 2013). A.2.3 M ODEL SIZE For all the models used in this paper, the size of a hidden layer nis 1000, the word embedding dimensionality mis 620 and the size of the maxout hidden layer in the deep output lis 500. The number of hidden units in the alignment model n0is 1000. B T RAINING PROCEDURE B.1 P ARAMETER INITIALIZATION We initialized the recurrent weight matrices U;Uz;Ur; U; Uz; Ur; !U; !Uzand !Uras random orthogonal matrices. For WaandUa, we initialized them by sampling each element from the Gaussian

1409.0473

Neural Machine Translation by Jointly Learning to Align and Translate

Neural machine translation is a recently proposed approach to machine translation. Unlike the traditional statistical machine translation, the neural machine translation aims at building a single neural network that can be jointly tuned to maximize the translation performance. The models proposed recently for neural machine translation often belong to a family of encoder-decoders and consists of an encoder that encodes a source sentence into a fixed-length vector from which a decoder generates a translation. In this paper, we conjecture that the use of a fixed-length vector is a bottleneck in improving the performance of this basic encoder-decoder architecture, and propose to extend this by allowing a model to automatically (soft-)search for parts of a source sentence that are relevant to predicting a target word, without having to form these parts as a hard segment explicitly. With this new approach, we achieve a translation performance comparable to the existing state-of-the-art phrase-based system on the task of English-to-French translation. Furthermore, qualitative analysis reveals that the (soft-)alignments found by the model agree well with our intuition.

http://arxiv.org/pdf/1409.0473

Accepted at ICLR 2015 as oral presentation

cs.CL

20140901

20160519

1409.0473

48

distribution of mean 0and variance 0:0012. All the elements of Vaand all the bias vectors were initialized to zero. Any other weight matrix was initialized by sampling from the Gaussian distribution of mean 0and variance 0:012. B.2 T RAINING We used the stochastic gradient descent (SGD) algorithm. Adadelta (Zeiler, 2012) was used to automatically adapt the learning rate of each parameter ( = 106and= 0:95). We explicitly 14 Published as a conference paper at ICLR 2015 normalized the L2-norm of the gradient of the cost function each time to be at most a predefined threshold of 1, when the norm was larger than the threshold (Pascanu et al. , 2013b). Each SGD update direction was computed with a minibatch of 80 sentences. At each update our implementation requires time proportional to the length of the longest sentence in a minibatch. Hence, to minimize the waste of computation, before every 20-th update, we retrieved 1600 sentence pairs, sorted them according to the lengths and split them into 20 minibatches. The training data was shuffled once before training and was traversed sequentially in this manner. In Tables 2 we present the statistics related to training all the models used in the experiments.

1409.0473

Neural Machine Translation by Jointly Learning to Align and Translate

Neural machine translation is a recently proposed approach to machine translation. Unlike the traditional statistical machine translation, the neural machine translation aims at building a single neural network that can be jointly tuned to maximize the translation performance. The models proposed recently for neural machine translation often belong to a family of encoder-decoders and consists of an encoder that encodes a source sentence into a fixed-length vector from which a decoder generates a translation. In this paper, we conjecture that the use of a fixed-length vector is a bottleneck in improving the performance of this basic encoder-decoder architecture, and propose to extend this by allowing a model to automatically (soft-)search for parts of a source sentence that are relevant to predicting a target word, without having to form these parts as a hard segment explicitly. With this new approach, we achieve a translation performance comparable to the existing state-of-the-art phrase-based system on the task of English-to-French translation. Furthermore, qualitative analysis reveals that the (soft-)alignments found by the model agree well with our intuition.

http://arxiv.org/pdf/1409.0473

Accepted at ICLR 2015 as oral presentation

cs.CL

20140901

20160519

1409.0473

49

In Tables 2 we present the statistics related to training all the models used in the experiments. C T RANSLATIONS OF LONG SENTENCES Source An admitting privilege is the right of a doctor to admit a patient to a hospital or a medical centre to carry out a diagnosis or a procedure, based on his status as a health care worker at a hospital. Reference Le privil `ege d’admission est le droit d’un m ´edecin, en vertu de son statut de membre soignant d’un h ˆopital, d’admettre un patient dans un h ˆopital ou un centre m ´edical afin d’y d ´elivrer un diagnostic ou un traitement. RNNenc-50 Un privil `ege d’admission est le droit d’un m ´edecin de reconna ˆıtre un patient `a l’h ˆopital ou un centre m ´edical d’un diagnostic ou de prendre un diagnostic en fonction de son ´etat de sant ´e.

1409.0473

Neural Machine Translation by Jointly Learning to Align and Translate

Neural machine translation is a recently proposed approach to machine translation. Unlike the traditional statistical machine translation, the neural machine translation aims at building a single neural network that can be jointly tuned to maximize the translation performance. The models proposed recently for neural machine translation often belong to a family of encoder-decoders and consists of an encoder that encodes a source sentence into a fixed-length vector from which a decoder generates a translation. In this paper, we conjecture that the use of a fixed-length vector is a bottleneck in improving the performance of this basic encoder-decoder architecture, and propose to extend this by allowing a model to automatically (soft-)search for parts of a source sentence that are relevant to predicting a target word, without having to form these parts as a hard segment explicitly. With this new approach, we achieve a translation performance comparable to the existing state-of-the-art phrase-based system on the task of English-to-French translation. Furthermore, qualitative analysis reveals that the (soft-)alignments found by the model agree well with our intuition.

http://arxiv.org/pdf/1409.0473

Accepted at ICLR 2015 as oral presentation

cs.CL

20140901

20160519

1409.0473

50

centre m ´edical d’un diagnostic ou de prendre un diagnostic en fonction de son ´etat de sant ´e. RNNsearch-50 Un privil `ege d’admission est le droit d’un m ´edecin d’admettre un patient `a un h ˆopital ou un centre m ´edical pour effectuer un diagnostic ou une proc ´edure, selon son statut de travailleur des soins de sant ´e`a l’h ˆopital. Google TranslateUn privil `ege admettre est le droit d’un m ´edecin d’admettre un patient dans un h ˆopital ou un centre m ´edical pour effectuer un diagnostic ou une proc ´edure, fond ´ee sur sa situation en tant que travailleur de soins de sant ´e dans un h ˆopital. Source This kind of experience is part of Disney’s efforts to ”extend the lifetime of its series and build new relationships with audiences via digital platforms that are becoming ever more important,” he added.

1409.0473

Neural Machine Translation by Jointly Learning to Align and Translate

Neural machine translation is a recently proposed approach to machine translation. Unlike the traditional statistical machine translation, the neural machine translation aims at building a single neural network that can be jointly tuned to maximize the translation performance. The models proposed recently for neural machine translation often belong to a family of encoder-decoders and consists of an encoder that encodes a source sentence into a fixed-length vector from which a decoder generates a translation. In this paper, we conjecture that the use of a fixed-length vector is a bottleneck in improving the performance of this basic encoder-decoder architecture, and propose to extend this by allowing a model to automatically (soft-)search for parts of a source sentence that are relevant to predicting a target word, without having to form these parts as a hard segment explicitly. With this new approach, we achieve a translation performance comparable to the existing state-of-the-art phrase-based system on the task of English-to-French translation. Furthermore, qualitative analysis reveals that the (soft-)alignments found by the model agree well with our intuition.

http://arxiv.org/pdf/1409.0473

Accepted at ICLR 2015 as oral presentation

cs.CL

20140901

20160519

1409.0473

51

new relationships with audiences via digital platforms that are becoming ever more important,” he added. Reference Ce type d’exp ´erience entre dans le cadre des efforts de Disney pour ” ´etendre la dur ´ee de vie de ses s ´eries et construire de nouvelles relations avec son public gr ˆace`a des plateformes num´eriques qui sont de plus en plus importantes”, a-t-il ajout ´e. RNNenc-50 Ce type d’exp ´erience fait partie des initiatives du Disney pour ”prolonger la dur ´ee de vie de ses nouvelles et de d ´evelopper des liens avec les lecteurs num ´eriques qui deviennent plus complexes. RNNsearch-50 Ce genre d’exp ´erience fait partie des efforts de Disney pour ”prolonger la dur ´ee de vie de ses s´eries et cr ´eer de nouvelles relations avec des publics via des plateformes num ´eriques de plus en plus importantes”, a-t-il ajout ´e. Google

1409.0473

Neural Machine Translation by Jointly Learning to Align and Translate

Neural machine translation is a recently proposed approach to machine translation. Unlike the traditional statistical machine translation, the neural machine translation aims at building a single neural network that can be jointly tuned to maximize the translation performance. The models proposed recently for neural machine translation often belong to a family of encoder-decoders and consists of an encoder that encodes a source sentence into a fixed-length vector from which a decoder generates a translation. In this paper, we conjecture that the use of a fixed-length vector is a bottleneck in improving the performance of this basic encoder-decoder architecture, and propose to extend this by allowing a model to automatically (soft-)search for parts of a source sentence that are relevant to predicting a target word, without having to form these parts as a hard segment explicitly. With this new approach, we achieve a translation performance comparable to the existing state-of-the-art phrase-based system on the task of English-to-French translation. Furthermore, qualitative analysis reveals that the (soft-)alignments found by the model agree well with our intuition.

http://arxiv.org/pdf/1409.0473

Accepted at ICLR 2015 as oral presentation

cs.CL

20140901

20160519

1409.0473

52

en plus importantes”, a-t-il ajout ´e. Google TranslateCe genre d’exp ´erience fait partie des efforts de Disney `a “´etendre la dur ´ee de vie de sa s ´erie et construire de nouvelles relations avec le public par le biais des plates-formes num ´eriques qui deviennent de plus en plus important”, at-il ajout ´e. Source In a press conference on Thursday, Mr Blair stated that there was nothing in this video that might constitute a ”reasonable motive” that could lead to criminal charges being brought against the mayor. Reference En conf ´erence de presse, jeudi, M. Blair a affirm ´e qu’il n’y avait rien dans cette vid ´eo qui puisse constituer des ”motifs raisonnables” pouvant mener au d ´epˆot d’une accusation criminelle contre le maire.

1409.0473

Neural Machine Translation by Jointly Learning to Align and Translate

Neural machine translation is a recently proposed approach to machine translation. Unlike the traditional statistical machine translation, the neural machine translation aims at building a single neural network that can be jointly tuned to maximize the translation performance. The models proposed recently for neural machine translation often belong to a family of encoder-decoders and consists of an encoder that encodes a source sentence into a fixed-length vector from which a decoder generates a translation. In this paper, we conjecture that the use of a fixed-length vector is a bottleneck in improving the performance of this basic encoder-decoder architecture, and propose to extend this by allowing a model to automatically (soft-)search for parts of a source sentence that are relevant to predicting a target word, without having to form these parts as a hard segment explicitly. With this new approach, we achieve a translation performance comparable to the existing state-of-the-art phrase-based system on the task of English-to-French translation. Furthermore, qualitative analysis reveals that the (soft-)alignments found by the model agree well with our intuition.

http://arxiv.org/pdf/1409.0473

Accepted at ICLR 2015 as oral presentation

cs.CL

20140901

20160519

1409.0473

53

le maire. RNNenc-50 Lors de la conf ´erence de presse de jeudi, M. Blair a dit qu’il n’y avait rien dans cette vid ´eo qui pourrait constituer une ”motivation raisonnable” pouvant entra ˆıner des accusations criminelles port´ees contre le maire. RNNsearch-50 Lors d’une conf ´erence de presse jeudi, M. Blair a d ´eclar ´e qu’il n’y avait rien dans cette vid ´eo qui pourrait constituer un ”motif raisonnable” qui pourrait conduire `a des accusations criminelles contre le maire. Google TranslateLors d’une conf ´erence de presse jeudi, M. Blair a d ´eclar ´e qu’il n’y avait rien dans cette vido qui pourrait constituer un ”motif raisonnable” qui pourrait mener `a des accusations criminelles portes contre le maire.

1409.0473

Neural Machine Translation by Jointly Learning to Align and Translate

Neural machine translation is a recently proposed approach to machine translation. Unlike the traditional statistical machine translation, the neural machine translation aims at building a single neural network that can be jointly tuned to maximize the translation performance. The models proposed recently for neural machine translation often belong to a family of encoder-decoders and consists of an encoder that encodes a source sentence into a fixed-length vector from which a decoder generates a translation. In this paper, we conjecture that the use of a fixed-length vector is a bottleneck in improving the performance of this basic encoder-decoder architecture, and propose to extend this by allowing a model to automatically (soft-)search for parts of a source sentence that are relevant to predicting a target word, without having to form these parts as a hard segment explicitly. With this new approach, we achieve a translation performance comparable to the existing state-of-the-art phrase-based system on the task of English-to-French translation. Furthermore, qualitative analysis reveals that the (soft-)alignments found by the model agree well with our intuition.

http://arxiv.org/pdf/1409.0473

Accepted at ICLR 2015 as oral presentation

cs.CL

20140901

20160519

1409.0473

54

qui pourrait constituer un ”motif raisonnable” qui pourrait mener `a des accusations criminelles portes contre le maire. Table 3: The translations generated by RNNenc-50 and RNNsearch-50 from long source sentences (30 words or more) selected from the test set. For each source sentence, we also show the goldstandard translation. The translations by Google Translate were made on 27 August 2014. 15

1409.0473

Neural Machine Translation by Jointly Learning to Align and Translate

Neural machine translation is a recently proposed approach to machine translation. Unlike the traditional statistical machine translation, the neural machine translation aims at building a single neural network that can be jointly tuned to maximize the translation performance. The models proposed recently for neural machine translation often belong to a family of encoder-decoders and consists of an encoder that encodes a source sentence into a fixed-length vector from which a decoder generates a translation. In this paper, we conjecture that the use of a fixed-length vector is a bottleneck in improving the performance of this basic encoder-decoder architecture, and propose to extend this by allowing a model to automatically (soft-)search for parts of a source sentence that are relevant to predicting a target word, without having to form these parts as a hard segment explicitly. With this new approach, we achieve a translation performance comparable to the existing state-of-the-art phrase-based system on the task of English-to-French translation. Furthermore, qualitative analysis reveals that the (soft-)alignments found by the model agree well with our intuition.

http://arxiv.org/pdf/1409.0473

Accepted at ICLR 2015 as oral presentation

cs.CL

20140901

20160519

1409.1556

0

arXiv:1409.1556v6 [cs.CV] 10 Apr 2015Publishedasa conferencepaperat ICLR2015 VERYDEEPCONVOLUTIONAL NETWORKS FORLARGE-SCALEIMAGERECOGNITION KarenSimonyan∗& AndrewZisserman+ VisualGeometryGroup,DepartmentofEngineeringScience, UniversityofOxford {karen,az }@robots.ox.ac.uk ABSTRACT In this work we investigate the effect of the convolutional n etwork depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with verysmall( 3×3)convolutionfilters,whichshowsthatasignificantimprove ment on the prior-art configurations can be achieved by pushing th e depth to 16–19 weight layers. These findings were the basis of our ImageNet C hallenge 2014 submission,whereourteamsecuredthefirstandthesecondpl acesinthelocalisation and classification tracks respectively. We also show th at our representations generalise well to other datasets, where they achieve state -of-the-art results. We

1409.1556

Very Deep Convolutional Networks for Large-Scale Image Recognition

In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

http://arxiv.org/pdf/1409.1556

cs.CV

20140904

20150410

1409.1556

1

generalise well to other datasets, where they achieve state -of-the-art results. We have made our two best-performingConvNet models publicly a vailable to facilitate furtherresearchontheuse ofdeepvisualrepresentati onsincomputervision. 1 INTRODUCTION Convolutional networks (ConvNets) have recently enjoyed a great success in large-scale image and video recognition (Krizhevskyetal., 2012; Zeiler& Fergus, 2013; Sermanetet al., 2014; Simonyan& Zisserman, 2014) which has become possible due to the large public image repositories,suchasImageNet(Denget al.,2009),andhigh-perform ancecomputingsystems,suchasGPUs orlarge-scaledistributedclusters(Deanet al., 2012). In particular,animportantroleintheadvance ofdeepvisualrecognitionarchitectureshasbeenplayedby theImageNetLarge-ScaleVisualRecognition Challenge (ILSVRC) (Russakovskyet al., 2014), whic h has served as a testbed for a few

1409.1556

Very Deep Convolutional Networks for Large-Scale Image Recognition

In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

http://arxiv.org/pdf/1409.1556

cs.CV

20140904

20150410

1409.1556

2

generationsof large-scale image classification systems, f rom high-dimensionalshallow feature encodings(Perronninetal.,2010)(thewinnerofILSVRC-2011 )todeepConvNets(Krizhevskyet al., 2012)(thewinnerofILSVRC-2012). With ConvNets becoming more of a commodity in the computer vi sion field, a number of attempts have been made to improve the original architecture o f Krizhevskyet al. (2012) in a bid to achieve better accuracy. For instance, the best-perf orming submissions to the ILSVRC2013 (Zeiler&Fergus, 2013; Sermanetetal., 2014) utilised smaller receptive window size and smaller stride of the first convolutional layer. Another lin e of improvements dealt with training and testing the networks densely over the whole image and ove r multiple scales (Sermanetet al., 2014; Howard, 2014). In this paper, we address another impor tant aspect of ConvNet architecture design–itsdepth. Tothisend,we fixotherparametersofthea rchitecture,andsteadilyincreasethe depth of the network by adding more convolutionallayers, wh ich is feasible due to the use of very

1409.1556

Very Deep Convolutional Networks for Large-Scale Image Recognition

In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

http://arxiv.org/pdf/1409.1556

cs.CV

20140904

20150410

1409.1556

3

depth of the network by adding more convolutionallayers, wh ich is feasible due to the use of very small (3×3)convolutionfiltersinall layers. As a result, we come up with significantly more accurate ConvN et architectures, which not only achieve the state-of-the-art accuracy on ILSVRC classifica tion and localisation tasks, but are also applicabletootherimagerecognitiondatasets,wherethey achieveexcellentperformanceevenwhen usedasa partofa relativelysimple pipelines(e.g.deepfea turesclassified byalinearSVM without fine-tuning). We havereleasedourtwobest-performingmode ls1tofacilitatefurtherresearch. The rest of the paper is organised as follows. In Sect. 2, we de scribe our ConvNet configurations. The details of the image classification trainingand evaluat ionare then presented in Sect. 3, and the ∗current affiliation: Google DeepMind+current affiliation: Universityof Oxfordand Google DeepMi nd 1http://www.robots.ox.ac.uk/ ˜vgg/research/very_deep/ 1

1409.1556

Very Deep Convolutional Networks for Large-Scale Image Recognition

In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

http://arxiv.org/pdf/1409.1556

cs.CV

20140904

20150410

1409.1556

4

1http://www.robots.ox.ac.uk/ ˜vgg/research/very_deep/ 1 Publishedasa conferencepaperat ICLR2015 configurations are compared on the ILSVRC classification tas k in Sect. 4. Sect. 5 concludes the paper. For completeness,we also describeand assess our ILS VRC-2014object localisationsystem inAppendixA,anddiscussthegeneralisationofverydeepfe aturestootherdatasetsinAppendixB. Finally,AppendixCcontainsthelist ofmajorpaperrevisio ns. 2 CONVNETCONFIGURATIONS To measure the improvement brought by the increased ConvNet depth in a fair setting, all our ConvNet layer configurations are designed using the same pri nciples, inspired by Ciresan etal. (2011); Krizhevskyet al. (2012). In this section, we first de scribe a generic layout of our ConvNet configurations(Sect.2.1)andthendetailthespecificconfig urationsusedintheevaluation(Sect.2.2).

1409.1556

Very Deep Convolutional Networks for Large-Scale Image Recognition

In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

http://arxiv.org/pdf/1409.1556

cs.CV

20140904

20150410

1409.1556

5

Ourdesignchoicesarethendiscussedandcomparedtothepri orart inSect. 2.3. 2.1 A RCHITECTURE During training, the input to our ConvNets is a fixed-size 224×224RGB image. The only preprocessingwedoissubtractingthemeanRGBvalue,computed onthetrainingset,fromeachpixel. Theimageispassedthroughastackofconvolutional(conv.) layers,whereweusefilterswithavery small receptive field: 3×3(which is the smallest size to capture the notion of left/rig ht, up/down, center). In one of the configurationswe also utilise 1×1convolutionfilters, which can be seen as a linear transformationof the input channels (followed by n on-linearity). The convolutionstride is fixedto1pixel;thespatialpaddingofconv.layerinputissuchthatt hespatialresolutionispreserved afterconvolution,i.e. the paddingis 1pixel for3×3conv.layers. Spatial poolingis carriedoutby

1409.1556

Very Deep Convolutional Networks for Large-Scale Image Recognition

In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

http://arxiv.org/pdf/1409.1556

cs.CV

20140904

20150410

1409.1556

6

afterconvolution,i.e. the paddingis 1pixel for3×3conv.layers. Spatial poolingis carriedoutby fivemax-poolinglayers,whichfollowsomeoftheconv.layer s(notalltheconv.layersarefollowed bymax-pooling). Max-poolingisperformedovera 2×2pixelwindow,withstride 2. Astackofconvolutionallayers(whichhasadifferentdepth indifferentarchitectures)isfollowedby three Fully-Connected(FC) layers: the first two have4096ch annelseach,the thirdperforms1000way ILSVRC classification and thus contains1000channels(o ne foreach class). The final layer is thesoft-maxlayer. Theconfigurationofthefullyconnected layersis thesameinall networks. Allhiddenlayersareequippedwiththerectification(ReLU( Krizhevskyetal.,2012))non-linearity. We note that none of our networks (except for one) contain Loc al Response Normalisation (LRN) normalisation (Krizhevskyet al., 2012): as will be sh own in Sect. 4, such normalisation

1409.1556

Very Deep Convolutional Networks for Large-Scale Image Recognition

In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

http://arxiv.org/pdf/1409.1556

cs.CV

20140904

20150410

1409.1556

7

(LRN) normalisation (Krizhevskyet al., 2012): as will be sh own in Sect. 4, such normalisation does not improve the performance on the ILSVRC dataset, but l eads to increased memory consumption and computation time. Where applicable, the param eters for the LRN layer are those of(Krizhevskyetal., 2012). 2.2 C ONFIGURATIONS The ConvNet configurations, evaluated in this paper, are out lined in Table 1, one per column. In the following we will refer to the nets by their names (A–E). A ll configurationsfollow the generic design presented in Sect. 2.1, and differ only in the depth: f rom 11 weight layers in the network A (8conv.and3FClayers)to19weightlayersinthenetworkE(1 6conv.and3FClayers). Thewidth of conv.layers (the number of channels) is rather small, sta rting from 64in the first layer and then increasingbyafactorof 2aftereachmax-poolinglayer,untilit reaches 512. In Table 2 we reportthe numberof parametersfor each configur ation. In spite of a large depth, the

1409.1556

Very Deep Convolutional Networks for Large-Scale Image Recognition

In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

http://arxiv.org/pdf/1409.1556

cs.CV

20140904

20150410

1409.1556

8

In Table 2 we reportthe numberof parametersfor each configur ation. In spite of a large depth, the numberof weights in our netsis not greater thanthe numberof weightsin a moreshallow net with largerconv.layerwidthsandreceptivefields(144Mweights in(Sermanetet al., 2014)). 2.3 D ISCUSSION Our ConvNet configurations are quite different from the ones used in the top-performing entries of the ILSVRC-2012 (Krizhevskyetal., 2012) and ILSVRC-201 3 competitions (Zeiler& Fergus, 2013;Sermanetet al.,2014). Ratherthanusingrelativelyl argereceptivefieldsinthefirstconv.layers(e.g.11×11withstride 4in(Krizhevskyet al.,2012),or 7×7withstride 2in(Zeiler& Fergus, 2013; Sermanetet al., 2014)), we use very small 3×3receptive fields throughout the whole net, whichareconvolvedwiththeinputateverypixel(withstrid e1). Itiseasytoseethatastackoftwo

1409.1556

Very Deep Convolutional Networks for Large-Scale Image Recognition

In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

http://arxiv.org/pdf/1409.1556

cs.CV

20140904

20150410

1409.1556

9

whichareconvolvedwiththeinputateverypixel(withstrid e1). Itiseasytoseethatastackoftwo 3×3conv.layers(withoutspatialpoolinginbetween)hasaneff ectivereceptivefieldof 5×5;three 2 Publishedasa conferencepaperat ICLR2015 Table 1:ConvNet configurations (shown in columns). The depth of the configurations increase s fromtheleft(A)totheright(E),asmorelayersareadded(th eaddedlayersareshowninbold). The convolutional layer parameters are denoted as “conv /an}bracketle{treceptive field size /an}bracketri}ht-/an}bracketle{tnumber of channels /an}bracketri}ht”. TheReLU activationfunctionisnotshownforbrevity. ConvNet Configuration A A-LRN B C D E 11weight 11weight 13 weight 16weight 16weight 19 weight layers layers layers layers layers layers input (224×224RGBimage) conv3-64 conv3-64 conv3-64 conv3-64 conv3-64 conv3-64

1409.1556

Very Deep Convolutional Networks for Large-Scale Image Recognition

In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

http://arxiv.org/pdf/1409.1556

cs.CV

20140904

20150410

1409.1556

10

input (224×224RGBimage) conv3-64 conv3-64 conv3-64 conv3-64 conv3-64 conv3-64 LRN conv3-64 conv3-64 conv3-64 conv3-64 maxpool conv3-128 conv3-128 conv3-128 conv3-128 conv3-128 conv3-128 conv3-128 conv3-128 conv3-128 conv3-128 maxpool conv3-256 conv3-256 conv3-256 conv3-256 conv3-256 conv3-256 conv3-256 conv3-256 conv3-256 conv3-256 conv3-256 conv3-256 conv1-256 conv3-256 conv3-256 conv3-256 maxpool conv3-512 conv3-512 conv3-512 conv3-512 conv3-512 conv3-512 conv3-512 conv3-512 conv3-512 conv3-512 conv3-512 conv3-512 conv1-512 conv3-512 conv3-512 conv3-512 maxpool conv3-512 conv3-512 conv3-512 conv3-512 conv3-512 conv3-512 conv3-512 conv3-512 conv3-512 conv3-512 conv3-512 conv3-512

1409.1556

Very Deep Convolutional Networks for Large-Scale Image Recognition

In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

http://arxiv.org/pdf/1409.1556

cs.CV

20140904

20150410

1409.1556

11

conv3-512 conv3-512 conv3-512 conv3-512 conv3-512 conv3-512 conv1-512 conv3-512 conv3-512 conv3-512 maxpool FC-4096 FC-4096 FC-1000 soft-max Table2:Number ofparameters (inmillions). Network A,A-LRN BCDE Number of parameters 133 133134138144 such layers have a 7×7effectivereceptive field. So what have we gainedby using, fo r instance, a stackofthree 3×3conv.layersinsteadofasingle 7×7layer? First,weincorporatethreenon-linear rectification layers instead of a single one, which makes the decision functionmore discriminative. Second, we decrease the number of parameters: assuming that both the input and the output of a three-layer 3×3convolutionstack has Cchannels,the stack is parametrisedby 3/parenleftbig 32C2/parenrightbig = 27C2 weights; at the same time, a single 7×7conv. layer would require 72C2= 49C2parameters, i.e.

1409.1556

Very Deep Convolutional Networks for Large-Scale Image Recognition

In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

http://arxiv.org/pdf/1409.1556

cs.CV

20140904

20150410

1409.1556

12

= 27C2 weights; at the same time, a single 7×7conv. layer would require 72C2= 49C2parameters, i.e. 81%more. Thiscan be seen as imposinga regularisationon the 7×7conv.filters, forcingthemto haveadecompositionthroughthe 3×3filters(withnon-linearityinjectedin between). The incorporation of 1×1conv. layers (configuration C, Table 1) is a way to increase th e nonlinearity of the decision function without affecting the re ceptive fields of the conv. layers. Even thoughinourcasethe 1×1convolutionisessentiallyalinearprojectionontothespa ceofthesame dimensionality(thenumberofinputandoutputchannelsist hesame),anadditionalnon-linearityis introducedbytherectificationfunction. Itshouldbenoted that1×1conv.layershaverecentlybeen utilisedin the“NetworkinNetwork”architectureofLinet a l.(2014). Small-size convolution filters have been previously used by Ciresan etal. (2011), but their nets

1409.1556

Very Deep Convolutional Networks for Large-Scale Image Recognition

In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

http://arxiv.org/pdf/1409.1556

cs.CV

20140904

20150410

1409.1556

13

Small-size convolution filters have been previously used by Ciresan etal. (2011), but their nets are significantly less deep than ours, and they did not evalua te on the large-scale ILSVRC dataset. Goodfellowet al. (2014) applied deep ConvNets ( 11weight layers) to the task of street number recognition, and showed that the increased de pth led to better performance. GoogLeNet(Szegedyet al., 2014), a top-performingentryof the ILSVRC-2014classification task, was developed independentlyof our work, but is similar in th at it is based on very deep ConvNets 3 Publishedasa conferencepaperat ICLR2015 (22 weight layers) and small convolution filters (apart from 3×3, they also use 1×1and5×5 convolutions). Their network topology is, however, more co mplex than ours, and the spatial resolution of the feature maps is reduced more aggressively in th e first layers to decrease the amount of computation. As will be shown in Sect. 4.5, our model is out performing that of Szegedyetal.

1409.1556

Very Deep Convolutional Networks for Large-Scale Image Recognition

In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

http://arxiv.org/pdf/1409.1556

cs.CV

20140904

20150410

1409.1556

14

of computation. As will be shown in Sect. 4.5, our model is out performing that of Szegedyetal. (2014)intermsofthesingle-networkclassificationaccura cy. 3 CLASSIFICATION FRAMEWORK In the previous section we presented the details of our netwo rk configurations. In this section, we describethe detailsofclassificationConvNettrainingand evaluation. 3.1 T RAINING The ConvNet training procedure generally follows Krizhevs kyetal. (2012) (except for sampling theinputcropsfrommulti-scaletrainingimages,asexplai nedlater). Namely,thetrainingiscarried out by optimising the multinomial logistic regression obje ctive using mini-batch gradient descent (based on back-propagation(LeCunet al., 1989)) with momen tum. The batch size was set to 256, momentum to 0.9. The training was regularised by weight decay (the L2penalty multiplier set to 5·10−4)anddropoutregularisationforthefirsttwofully-connect edlayers(dropoutratiosetto 0.5).

1409.1556

Very Deep Convolutional Networks for Large-Scale Image Recognition

In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

http://arxiv.org/pdf/1409.1556

cs.CV

20140904

20150410

1409.1556

15

5·10−4)anddropoutregularisationforthefirsttwofully-connect edlayers(dropoutratiosetto 0.5). Thelearningrate wasinitially setto 10−2,andthendecreasedbyafactorof 10whenthevalidation set accuracy stopped improving. In total, the learning rate was decreased 3 times, and the learning was stopped after 370K iterations (74 epochs). We conjecture that in spite of the l arger number of parametersandthegreaterdepthofournetscomparedto(Kri zhevskyetal.,2012),thenetsrequired lessepochstoconvergedueto(a)implicitregularisationi mposedbygreaterdepthandsmallerconv. filter sizes; (b)pre-initialisationofcertainlayers. The initialisation of the networkweightsis important,sin ce bad initialisation can stall learningdue to the instability of gradient in deep nets. To circumvent th is problem, we began with training the configurationA (Table 1), shallow enoughto be trained wi th randominitialisation. Then,when

1409.1556

Very Deep Convolutional Networks for Large-Scale Image Recognition

In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

http://arxiv.org/pdf/1409.1556

cs.CV

20140904

20150410

1409.1556

16

the configurationA (Table 1), shallow enoughto be trained wi th randominitialisation. Then,when trainingdeeperarchitectures,weinitialisedthefirstfou rconvolutionallayersandthelastthreefullyconnectedlayerswiththelayersofnetA(theintermediatel ayerswereinitialisedrandomly). Wedid notdecreasethelearningrateforthepre-initialisedlaye rs,allowingthemtochangeduringlearning. For random initialisation (where applicable), we sampled t he weights from a normal distribution with thezeromeanand 10−2variance. The biaseswere initialisedwith zero. It isworth notingthat after the paper submission we found that it is possible to ini tialise the weights without pre-training byusingthe randominitialisationprocedureofGlorot&Ben gio(2010). Toobtainthefixed-size 224×224ConvNetinputimages,theywererandomlycroppedfromresca led training images (one crop per image per SGD iteration). To fu rther augment the training set, the cropsunderwentrandomhorizontalflippingandrandomRGBco lourshift(Krizhevskyet al.,2012). Trainingimagerescalingisexplainedbelow.

1409.1556

Very Deep Convolutional Networks for Large-Scale Image Recognition

In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

http://arxiv.org/pdf/1409.1556

cs.CV

20140904

20150410

1409.1556

17

Trainingimagerescalingisexplainedbelow. Training image size. LetSbe the smallest side of an isotropically-rescaledtraining image, from which the ConvNet input is cropped (we also refer to Sas the training scale). While the crop size is fixed to 224×224, in principle Scan take on any value not less than 224: forS= 224the crop will capture whole-image statistics, completely spanning the smallest side of a training image; for S≫224thecropwillcorrespondtoasmallpartoftheimage,contain ingasmallobjectoranobject part. We considertwoapproachesforsettingthetrainingscale S. Thefirst istofix S,whichcorresponds to single-scale training (note that image content within th e sampled crops can still represent multiscale image statistics). In our experiments, we evaluated m odels trained at two fixed scales: S= 256(which has been widely used in the prior art (Krizhevskyet al ., 2012; Zeiler&Fergus, 2013; Sermanetet al., 2014)) and S= 384. Given a ConvNet configuration,we first trained the network

1409.1556

Very Deep Convolutional Networks for Large-Scale Image Recognition

In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

http://arxiv.org/pdf/1409.1556

cs.CV

20140904

20150410

1409.1556

18

Sermanetet al., 2014)) and S= 384. Given a ConvNet configuration,we first trained the network usingS= 256. To speed-up training of the S= 384network, it was initialised with the weights pre-trainedwith S= 256,andwe useda smallerinitiallearningrateof 10−3. The second approachto setting Sis multi-scale training, where each training image is indiv idually rescaled by randomly sampling Sfrom a certain range [Smin,Smax](we used Smin= 256and Smax= 512). Sinceobjectsinimagescanbeofdifferentsize,itisbene ficialtotakethisintoaccount duringtraining. Thiscanalso beseen astrainingset augmen tationbyscale jittering,wherea single 4 Publishedasa conferencepaperat ICLR2015 model is trained to recognise objects over a wide range of sca les. For speed reasons, we trained multi-scale models by fine-tuning all layers of a single-sca le model with the same configuration, pre-trainedwithfixed S= 384. 3.2 T ESTING

1409.1556

Very Deep Convolutional Networks for Large-Scale Image Recognition

In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

http://arxiv.org/pdf/1409.1556

cs.CV

20140904

20150410

1409.1556

19

pre-trainedwithfixed S= 384. 3.2 T ESTING Attest time,givena trainedConvNetandaninputimage,itis classified inthefollowingway. First, it is isotropically rescaled to a pre-defined smallest image side, denoted as Q(we also refer to it as the test scale). We note that Qis not necessarily equal to the training scale S(as we will show in Sect. 4, usingseveralvaluesof QforeachSleadsto improvedperformance). Then,the network is applied densely overthe rescaled test image in a way simil ar to (Sermanetet al., 2014). Namely, the fully-connected layers are first converted to convoluti onal layers (the first FC layer to a 7×7 conv. layer, the last two FC layers to 1×1conv. layers). The resulting fully-convolutional net is then applied to the whole (uncropped) image. The result is a c lass score map with the number of channels equal to the number of classes, and a variable spati al resolution, dependent on the input imagesize. Finally,toobtainafixed-sizevectorofclasssc oresfortheimage,theclassscoremapis

1409.1556

Very Deep Convolutional Networks for Large-Scale Image Recognition

In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

http://arxiv.org/pdf/1409.1556

cs.CV

20140904

20150410

1409.1556

20

imagesize. Finally,toobtainafixed-sizevectorofclasssc oresfortheimage,theclassscoremapis spatially averaged(sum-pooled). We also augmentthe test s et by horizontalflippingof the images; thesoft-maxclassposteriorsoftheoriginalandflippedima gesareaveragedtoobtainthefinalscores fortheimage. Since the fully-convolutionalnetwork is applied over the w hole image, there is no need to sample multiple crops at test time (Krizhevskyetal., 2012), which is less efficient as it requires network re-computationforeachcrop. Atthesametime,usingalarge setofcrops,asdonebySzegedyetal. (2014),canleadtoimprovedaccuracy,asit resultsin afiner samplingoftheinputimagecompared tothefully-convolutionalnet. Also,multi-cropevaluati oniscomplementarytodenseevaluationdue to different convolution boundary conditions: when applyi ng a ConvNet to a crop, the convolved feature mapsare paddedwith zeros, while in the case of dense evaluationthe paddingfor the same

1409.1556

Very Deep Convolutional Networks for Large-Scale Image Recognition

In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

http://arxiv.org/pdf/1409.1556

cs.CV

20140904

20150410

1409.1556

21

feature mapsare paddedwith zeros, while in the case of dense evaluationthe paddingfor the same crop naturally comes from the neighbouring parts of an image (due to both the convolutions and spatial pooling), which substantially increases the overa ll network receptive field, so more context iscaptured. Whilewebelievethatinpracticetheincreased computationtimeofmultiplecropsdoes notjustifythepotentialgainsinaccuracy,forreferencew ealsoevaluateournetworksusing 50crops perscale( 5×5regulargridwith 2flips),foratotalof 150cropsover 3scales,whichiscomparable to144cropsover 4scalesusedbySzegedyetal. (2014). 3.3 IMPLEMENTATION DETAILS OurimplementationisderivedfromthepubliclyavailableC ++ Caffetoolbox(Jia,2013)(branched out in December 2013), but contains a number of significant mo difications, allowing us to perform trainingandevaluationonmultipleGPUsinstalledinasing lesystem,aswellastrainandevaluateon full-size (uncropped) images at multiple scales (as descri bed above). Multi-GPU training exploits

1409.1556

Very Deep Convolutional Networks for Large-Scale Image Recognition

In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

http://arxiv.org/pdf/1409.1556

cs.CV

20140904

20150410

1409.1556

22

full-size (uncropped) images at multiple scales (as descri bed above). Multi-GPU training exploits data parallelism, and is carried out by splitting each batch of training images into several GPU batches, processed in parallel on each GPU. After the GPU bat ch gradientsare computed, they are averaged to obtain the gradient of the full batch. Gradient c omputation is synchronous across the GPUs, sothe resultisexactlythesame aswhentrainingona si ngleGPU. While more sophisticated methods of speeding up ConvNet tra ining have been recently proposed (Krizhevsky, 2014), which employmodeland data paral lelism for differentlayersof the net, wehavefoundthatourconceptuallymuchsimplerschemealre adyprovidesaspeedupof 3.75times on an off-the-shelf4-GPU system, as comparedto using a sing le GPU. On a system equippedwith fourNVIDIATitanBlackGPUs,trainingasinglenettook2–3w eeksdependingonthearchitecture. 4 CLASSIFICATION EXPERIMENTS Dataset. In this section, we present the image classification results achieved by the described

1409.1556

Very Deep Convolutional Networks for Large-Scale Image Recognition

In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

http://arxiv.org/pdf/1409.1556

cs.CV

20140904

20150410

1409.1556

23

4 CLASSIFICATION EXPERIMENTS Dataset. In this section, we present the image classification results achieved by the described ConvNetarchitecturesontheILSVRC-2012dataset(whichwa susedforILSVRC2012–2014challenges). The dataset includes images of 1000 classes, and is split into three sets: training ( 1.3M images), validation ( 50K images), and testing ( 100K images with held-out class labels). The classification performanceis evaluated using two measures: the top-1 and top-5 error. The former is a multi-class classification error, i.e. the proportion of in correctly classified images; the latter is the 5 Publishedasa conferencepaperat ICLR2015 main evaluation criterion used in ILSVRC, and is computed as the proportion of images such that theground-truthcategoryisoutsidethetop-5predictedca tegories. Forthemajorityofexperiments,weusedthevalidationseta sthetestset. Certainexperimentswere also carried out on the test set and submitted to the official I LSVRC server as a “VGG” team entry tothe ILSVRC-2014competition(Russakovskyet al., 2014).

1409.1556

Very Deep Convolutional Networks for Large-Scale Image Recognition

In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

http://arxiv.org/pdf/1409.1556

cs.CV

20140904

20150410

1409.1556

24

tothe ILSVRC-2014competition(Russakovskyet al., 2014). 4.1 SINGLESCALEEVALUATION We begin with evaluating the performanceof individual Conv Net models at a single scale with the layerconfigurationsdescribedin Sect. 2.2. The test images ize was set as follows: Q=Sforfixed S,andQ= 0.5(Smin+Smax)forjittered S∈[Smin,Smax]. Theresultsofareshownin Table3. First, we note that using local response normalisation (A-L RN network) does not improve on the model A without any normalisation layers. We thus do not empl oy normalisation in the deeper architectures(B–E). Second, we observe that the classification error decreases w ith the increased ConvNet depth: from 11 layers in A to 19 layers in E. Notably, in spite of the same de pth, the configuration C (which containsthree 1×1conv.layers),performsworsethantheconfigurationD,whic huses3×3conv. layersthroughoutthenetwork. Thisindicatesthatwhileth e additionalnon-linearitydoeshelp(Cis

1409.1556

Very Deep Convolutional Networks for Large-Scale Image Recognition

In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

http://arxiv.org/pdf/1409.1556

cs.CV

20140904

20150410

1409.1556

25

layersthroughoutthenetwork. Thisindicatesthatwhileth e additionalnon-linearitydoeshelp(Cis better than B), it is also important to capture spatial conte xt by using conv. filters with non-trivial receptive fields (D is better than C). The error rate of our arc hitecture saturates when the depth reaches19layers,butevendeepermodelsmightbebeneficialforlarger datasets. Wealsocompared the net B with a shallow net with five 5×5conv. layers, which was derived from B by replacing eachpairof 3×3conv. layerswithasingle 5×5conv. layer(whichhasthesamereceptivefieldas explained in Sect. 2.3). The top-1 error of the shallow net wa s measured to be 7%higher than that of B (on a center crop),which confirmsthat a deepnet with smal l filters outperformsa shallow net withlargerfilters. Finally, scale jittering at training time ( S∈[256;512] ) leads to significantly better results than

1409.1556

Very Deep Convolutional Networks for Large-Scale Image Recognition

In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

http://arxiv.org/pdf/1409.1556

cs.CV

20140904

20150410

1409.1556

26

Finally, scale jittering at training time ( S∈[256;512] ) leads to significantly better results than training on images with fixed smallest side ( S= 256orS= 384), even though a single scale is usedattesttime. Thisconfirmsthattrainingsetaugmentati onbyscalejitteringisindeedhelpfulfor capturingmulti-scaleimagestatistics. Table3:ConvNetperformanceatasingle testscale. ConvNet config. (Table 1) smallest image side top-1 val.error (%) top-5 val.error (%) train(S)test (Q) A 256 256 29.6 10.4 A-LRN 256 256 29.7 10.5 B 256 256 28.7 9.9 C256 256 28.1 9.4 384 384 28.1 9.3 [256;512] 384 27.3 8.8 D256 256 27.0 8.8 384 384 26.8 8.7 [256;512] 384 25.6 8.1 E256 256 27.3 9.0 384 384 26.9 8.7 [256;512] 384 25.5 8.0 4.2 M ULTI-SCALEEVALUATION

1409.1556

Very Deep Convolutional Networks for Large-Scale Image Recognition

In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

http://arxiv.org/pdf/1409.1556

cs.CV

20140904

20150410

1409.1556

27

384 384 26.9 8.7 [256;512] 384 25.5 8.0 4.2 M ULTI-SCALEEVALUATION HavingevaluatedtheConvNetmodelsatasinglescale,wenow assesstheeffectofscalejitteringat testtime. Itconsistsofrunningamodeloverseveralrescal edversionsofatestimage(corresponding to different values of Q), followed by averaging the resulting class posteriors. Co nsidering that a large discrepancy between training and testing scales lead s to a drop in performance, the models trained with fixed Swere evaluated over three test image sizes, close to the trai ning one: Q= {S−32,S,S+ 32}. At the same time, scale jittering at training time allows th e network to be appliedto a widerrangeofscales at test time,so the modeltr ainedwithvariable S∈[Smin;Smax] wasevaluatedoveralargerrangeofsizes Q={Smin,0.5(Smin+Smax),Smax}. 6 Publishedasa conferencepaperat ICLR2015 Theresults,presentedinTable4,indicatethatscalejitte ringattest timeleadstobetterperformance

1409.1556

Very Deep Convolutional Networks for Large-Scale Image Recognition

In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

http://arxiv.org/pdf/1409.1556

cs.CV

20140904

20150410

1409.1556

28

6 Publishedasa conferencepaperat ICLR2015 Theresults,presentedinTable4,indicatethatscalejitte ringattest timeleadstobetterperformance (as compared to evaluating the same model at a single scale, s hown in Table 3). As before, the deepest configurations(D and E) perform the best, and scale j ittering is better than training with a fixed smallest side S. Our best single-network performance on the validation set is24.8%/7.5% top-1/top-5error(highlightedinboldinTable4). Onthete stset,theconfigurationEachieves 7.3% top-5error. Table4:ConvNetperformanceatmultiple test scales. ConvNet config. (Table 1) smallest image side top-1val. error (%) top-5val. error (%) train(S)test(Q) B 256 224,256,288 28.2 9.6 C256 224,256,288 27.7 9.2 384 352,384,416 27.8 9.2 [256;512] 256,384,512 26.3 8.2 D256 224,256,288 26.6 8.6

1409.1556

Very Deep Convolutional Networks for Large-Scale Image Recognition

In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

http://arxiv.org/pdf/1409.1556

cs.CV

20140904

20150410

1409.1556

29

[256;512] 256,384,512 26.3 8.2 D256 224,256,288 26.6 8.6 384 352,384,416 26.5 8.6 [256;512] 256,384,512 24.8 7.5 E256 224,256,288 26.9 8.7 384 352,384,416 26.7 8.6 [256;512] 256,384,512 24.8 7.5 4.3 M ULTI-CROP EVALUATION In Table 5 we compare dense ConvNet evaluation with mult-cro p evaluation (see Sect. 3.2 for details). We also assess the complementarityof thetwo evalua tiontechniquesbyaveragingtheirsoftmax outputs. As can be seen, using multiple crops performs sl ightly better than dense evaluation, andthe two approachesareindeedcomplementary,astheir co mbinationoutperformseach ofthem. As noted above, we hypothesize that this is due to a different treatment of convolution boundary conditions. Table 5:ConvNetevaluationtechniques comparison. Inall experimentsthe trainingscale Swas sampledfrom [256;512] ,andthreetest scales Qwereconsidered: {256,384,512}.

1409.1556

Very Deep Convolutional Networks for Large-Scale Image Recognition

In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

http://arxiv.org/pdf/1409.1556

cs.CV

20140904

20150410

1409.1556

30

sampledfrom [256;512] ,andthreetest scales Qwereconsidered: {256,384,512}. ConvNet config. (Table 1) Evaluationmethod top-1 val. error(%) top-5 val. error (%) Ddense 24.8 7.5 multi-crop 24.6 7.5 multi-crop &dense 24.4 7.2 Edense 24.8 7.5 multi-crop 24.6 7.4 multi-crop &dense 24.4 7.1 4.4 C ONVNETFUSION Upuntilnow,weevaluatedtheperformanceofindividualCon vNetmodels. Inthispartoftheexperiments,wecombinetheoutputsofseveralmodelsbyaveragin gtheirsoft-maxclassposteriors. This improvesthe performancedueto complementarityof the mode ls, andwas used in the top ILSVRC submissions in 2012 (Krizhevskyet al., 2012) and 2013 (Zeil er&Fergus, 2013; Sermanetet al., 2014). The results are shown in Table 6. By the time of ILSVRC submiss ion we had only trained the single-scale networks, as well as a multi-scale model D (by fi ne-tuning only the fully-connected

1409.1556

Very Deep Convolutional Networks for Large-Scale Image Recognition

In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

http://arxiv.org/pdf/1409.1556

cs.CV

20140904

20150410

1409.1556

31

single-scale networks, as well as a multi-scale model D (by fi ne-tuning only the fully-connected layers rather than all layers). The resulting ensemble of 7 n etworks has 7.3%ILSVRC test error. After the submission, we considered an ensemble of only two b est-performing multi-scale models (configurations D and E), which reduced the test error to 7.0%using dense evaluation and 6.8% using combined dense and multi-crop evaluation. For refere nce, our best-performingsingle model achieves7.1%error(modelE, Table5). 4.5 C OMPARISON WITH THE STATE OF THE ART Finally, we compare our results with the state of the art in Ta ble 7. In the classification task of ILSVRC-2014 challenge (Russakovskyet al., 2014), our “VGG ” team secured the 2nd place with 7 Publishedasa conferencepaperat ICLR2015 Table6:Multiple ConvNetfusion results. Combined ConvNet modelsError top-1 val top-5val top-5test ILSVRCsubmission

1409.1556

Very Deep Convolutional Networks for Large-Scale Image Recognition

In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

http://arxiv.org/pdf/1409.1556

cs.CV

20140904

20150410

1409.1556

32

Table6:Multiple ConvNetfusion results. Combined ConvNet modelsError top-1 val top-5val top-5test ILSVRCsubmission (D/256/224,256,288), (D/384/352,384,416), (D/[256;512 ]/256,384,512) (C/256/224,256,288), (C/384/352,384,416) (E/256/224,256,288), (E/384/352,384,416)24.7 7.5 7.3 post-submission (D/[256;512]/256,384,512), (E/[256;512]/256,384,512) ,dense eval. 24.0 7.1 7.0 (D/[256;512]/256,384,512), (E/[256;512]/256,384,512) ,multi-crop 23.9 7.2 (D/[256;512]/256,384,512), (E/[256;512]/256,384,512) ,multi-crop &dense eval. 23.7 6.8 6.8 7.3%test errorusinganensembleof7 models. Afterthesubmissio n,we decreasedtheerrorrateto

1409.1556

Very Deep Convolutional Networks for Large-Scale Image Recognition

In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

http://arxiv.org/pdf/1409.1556

cs.CV

20140904

20150410

1409.1556

33

7.3%test errorusinganensembleof7 models. Afterthesubmissio n,we decreasedtheerrorrateto 6.8%usinganensembleof2models. As can be seen from Table 7, our very deep ConvNetssignificant ly outperformthe previousgeneration of models, which achieved the best results in the ILSVR C-2012 and ILSVRC-2013 competitions. Our result is also competitivewith respect to the cla ssification task winner(GoogLeNetwith 6.7%error) and substantially outperforms the ILSVRC-2013 winn ing submission Clarifai, which achieved 11.2%with outside training data and 11.7%without it. This is remarkable, considering that our best result is achievedby combiningjust two models – significantly less than used in most ILSVRC submissions. In terms of the single-net performance , our architecture achieves the best result (7.0%test error), outperforming a single GoogLeNet by 0.9%. Notably, we did not depart from the classical ConvNet architecture of LeCunetal. (198 9), but improved it by substantially increasingthedepth.

1409.1556

Very Deep Convolutional Networks for Large-Scale Image Recognition

In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

http://arxiv.org/pdf/1409.1556

cs.CV

20140904

20150410

1409.1556

34

from the classical ConvNet architecture of LeCunetal. (198 9), but improved it by substantially increasingthedepth. Table 7:Comparison with the state of the art in ILSVRC classification . Our methodis denoted as“VGG”.Onlytheresultsobtainedwithoutoutsidetrainin gdataarereported. Method top-1 val. error(%) top-5val. error (%) top-5testerror (%) VGG(2nets, multi-crop& dense eval.) 23.7 6.8 6.8 VGG(1net, multi-crop& dense eval.) 24.4 7.1 7.0 VGG(ILSVRCsubmission, 7nets, dense eval.) 24.7 7.5 7.3 GoogLeNet (Szegedy et al., 2014) (1net) - 7.9 GoogLeNet (Szegedy et al., 2014) (7nets) - 6.7 MSRA(He et al., 2014) (11nets) - - 8.1 MSRA(He et al., 2014) (1net) 27.9 9.1 9.1 Clarifai(Russakovsky et al., 2014) (multiplenets) - - 11.7

1409.1556

Very Deep Convolutional Networks for Large-Scale Image Recognition

In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

http://arxiv.org/pdf/1409.1556

cs.CV

20140904

20150410

1409.1556

35

Clarifai(Russakovsky et al., 2014) (multiplenets) - - 11.7 Clarifai(Russakovsky et al., 2014) (1net) - - 12.5 Zeiler& Fergus (Zeiler&Fergus, 2013) (6nets) 36.0 14.7 14.8 Zeiler& Fergus (Zeiler&Fergus, 2013) (1net) 37.5 16.0 16.1 OverFeat (Sermanetet al.,2014) (7nets) 34.0 13.2 13.6 OverFeat (Sermanetet al.,2014) (1net) 35.7 14.2 Krizhevsky et al.(Krizhevsky et al., 2012) (5nets) 38.1 16.4 16.4 Krizhevsky et al.(Krizhevsky et al., 2012) (1net) 40.7 18.2 5 CONCLUSION In this work we evaluated very deep convolutional networks ( up to 19 weight layers) for largescale image classification. It was demonstrated that the rep resentation depth is beneficial for the classificationaccuracy,andthatstate-of-the-artperfor manceontheImageNetchallengedatasetcan

1409.1556

Very Deep Convolutional Networks for Large-Scale Image Recognition

In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

http://arxiv.org/pdf/1409.1556

cs.CV

20140904

20150410

1409.1556

36

classificationaccuracy,andthatstate-of-the-artperfor manceontheImageNetchallengedatasetcan beachievedusingaconventionalConvNetarchitecture(LeC unet al.,1989;Krizhevskyet al.,2012) withsubstantiallyincreaseddepth. Intheappendix,weals oshowthatourmodelsgeneralisewellto a wide range of tasks and datasets, matchingor outperformin gmore complexrecognitionpipelines builtaroundlessdeepimagerepresentations. Ourresultsy etagainconfirmtheimportanceof depth invisualrepresentations. ACKNOWLEDGEMENTS ThisworkwassupportedbyERCgrantVisRecno.228180. Wegra tefullyacknowledgethesupport ofNVIDIACorporationwiththedonationoftheGPUsusedfort hisresearch. 8 Publishedasa conferencepaperat ICLR2015 REFERENCES Bell, S., Upchurch, P.,Snavely, N., and Bala, K. Material re cognition inthe wild withthe materials in context database. CoRR,abs/1412.0623, 2014.

1409.1556

Very Deep Convolutional Networks for Large-Scale Image Recognition

In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

http://arxiv.org/pdf/1409.1556

cs.CV

20140904

20150410

1409.1556

37

database. CoRR,abs/1412.0623, 2014. Chatfield, K., Simonyan, K., Vedaldi, A., and Zisserman, A. R eturn of the devil in the details: Delving deep intoconvolutional nets. In Proc.BMVC. ,2014. Cimpoi,M.,Maji,S.,andVedaldi,A. Deepconvolutionalfilt erbanksfortexturerecognitionandsegmentation. CoRR,abs/1411.6836, 2014. Ciresan, D. C., Meier, U., Masci, J., Gambardella, L. M., and Schmidhuber, J. Flexible, high performance convolutional neural networks for image classification. In IJCAI,pp. 1237–1242, 2011. Dean, J., Corrado, G., Monga, R., Chen, K., Devin, M., Mao, M. , Ranzato, M., Senior, A., Tucker, P., Yang, K.,Le,Q. V.,andNg, A.Y. Large scale distributeddeepnetwo rks. InNIPS,pp. 1232–1240, 2012.

1409.1556

Very Deep Convolutional Networks for Large-Scale Image Recognition

In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

http://arxiv.org/pdf/1409.1556

cs.CV

20140904

20150410

1409.1556

38

K.,Le,Q. V.,andNg, A.Y. Large scale distributeddeepnetwo rks. InNIPS,pp. 1232–1240, 2012. Deng, J., Dong, W., Socher, R., Li, L.-J., Li, K., and Fei-Fei , L. Imagenet: A large-scale hierarchical image database. In Proc.CVPR ,2009. Donahue,J.,Jia,Y.,Vinyals,O.,Hoffman,J.,Zhang,N.,Tz eng,E.,andDarrell,T.Decaf: Adeepconvolutional activation feature for generic visual recognition. CoRR,abs/1310.1531, 2013. Everingham, M., Eslami, S.M. A., Van Gool, L., Williams,C., Winn, J., and Zisserman, A. The Pascal visual object classes challenge: Aretrospective. IJCV,111(1):98–136, 2015. Fei-Fei, L., Fergus, R., and Perona, P. Learning generative visual models from few training examples: An incremental bayesian approach tested on 101 object categor ies. InIEEE CVPR Workshop of Generative Model BasedVision , 2004.

1409.1556

Very Deep Convolutional Networks for Large-Scale Image Recognition

In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

http://arxiv.org/pdf/1409.1556

cs.CV

20140904

20150410

1409.1556

39

incremental bayesian approach tested on 101 object categor ies. InIEEE CVPR Workshop of Generative Model BasedVision , 2004. Girshick, R. B., Donahue, J., Darrell, T., and Malik, J. Rich feature hierarchies for accurate object detection and semantic segmentation. CoRR,abs/1311.2524v5, 2014. PublishedinProc.CVPR,2014. Gkioxari, G.,Girshick, R.,and Malik, J. Actions and attrib utes from wholes and parts. CoRR,abs/1412.2604, 2014. Glorot, X. andBengio, Y. Understanding the difficultyof tra iningdeep feedforward neural networks. In Proc. AISTATS,volume 9, pp. 249–256, 2010. Goodfellow, I. J., Bulatov, Y., Ibarz, J., Arnoud, S., and Sh et, V. Multi-digit number recognition from street view imagery usingdeep convolutional neural networks. In Proc.ICLR ,2014. Griffin, G., Holub, A., and Perona, P. Caltech-256 object cat egory dataset. Technical Report 7694, California Institute of Technology, 2007.

1409.1556

Very Deep Convolutional Networks for Large-Scale Image Recognition

In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

http://arxiv.org/pdf/1409.1556

cs.CV

20140904

20150410

1409.1556

40

Institute of Technology, 2007. He, K., Zhang, X., Ren, S., and Sun, J. Spatial pyramid poolin g in deep convolutional networks for visual recognition. CoRR,abs/1406.4729v2, 2014. Hoai, M. Regularizedmax pooling forimage categorization. InProc. BMVC. ,2014. Howard, A.G. Someimprovements ondeepconvolutional neura l networkbasedimageclassification. In Proc. ICLR,2014. Jia, Y. Caffe: An open source convolutional architecture fo r fast feature embedding. http://caffe.berkeleyvision.org/ ,2013. Karpathy, A. and Fei-Fei, L. Deep visual-semantic alignmen ts for generating image descriptions. CoRR, abs/1412.2306, 2014. Kiros, R., Salakhutdinov, R., and Zemel, R. S. Unifying visu al-semantic embeddings with multimodal neural language models. CoRR,abs/1411.2539, 2014. Krizhevsky, A. One weirdtrickfor parallelizingconvoluti onal neural networks. CoRR,abs/1404.5997, 2014.

1409.1556

Very Deep Convolutional Networks for Large-Scale Image Recognition

In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

http://arxiv.org/pdf/1409.1556

cs.CV

20140904

20150410

1409.1556

41

Krizhevsky, A. One weirdtrickfor parallelizingconvoluti onal neural networks. CoRR,abs/1404.5997, 2014. Krizhevsky, A., Sutskever, I., and Hinton, G. E. ImageNet cl assification with deep convolutional neural networks. In NIPS,pp. 1106–1114, 2012. LeCun,Y.,Boser, B.,Denker, J.S.,Henderson, D.,Howard, R .E.,Hubbard, W.,andJackel, L.D. Backpropagationapplied tohandwrittenzipcode recognition. Neural Computation , 1(4):541–551, 1989. Lin,M., Chen, Q.,andYan, S. Networkinnetwork. In Proc.ICLR ,2014. Long, J., Shelhamer, E., and Darrell, T. Fully convolutiona l networks for semantic segmentation. CoRR, abs/1411.4038, 2014. Oquab, M., Bottou, L., Laptev, I., and Sivic, J. Learning and Transferring Mid-Level Image Representations using Convolutional Neural Networks. In Proc.CVPR ,2014.

1409.1556

Very Deep Convolutional Networks for Large-Scale Image Recognition

In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

http://arxiv.org/pdf/1409.1556

cs.CV

20140904

20150410

1409.1556

42

using Convolutional Neural Networks. In Proc.CVPR ,2014. Perronnin, F.,S´ anchez, J.,andMensink, T. Improving theF isherkernel forlarge-scale image classification. In Proc.ECCV ,2010. Razavian, A.,Azizpour, H.,Sullivan, J.,andCarlsson,S. C NNFeaturesoff-the-shelf: anAstounding Baseline for Recognition. CoRR,abs/1403.6382, 2014. 9 Publishedasa conferencepaperat ICLR2015 Russakovsky, O., Deng, J., Su, H., Krause, J., Satheesh, S., Ma, S., Huang, Z., Karpathy, A., Khosla, A., Bernstein, M., Berg, A. C., and Fei-Fei, L. ImageNet large sc ale visual recognition challenge. CoRR, abs/1409.0575, 2014. Sermanet,P.,Eigen, D.,Zhang, X.,Mathieu, M.,Fergus,R., andLeCun,Y. OverFeat: IntegratedRecognition, Localizationand Detectionusing Convolutional Networks. InProc.ICLR ,2014.

1409.1556

Very Deep Convolutional Networks for Large-Scale Image Recognition

In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

http://arxiv.org/pdf/1409.1556

cs.CV

20140904

20150410

1409.1556

43

Localizationand Detectionusing Convolutional Networks. InProc.ICLR ,2014. Simonyan, K. and Zisserman, A. Two-stream convolutional ne tworks for action recognition in videos. CoRR, abs/1406.2199, 2014. Published inProc.NIPS,2014. Szegedy, C., Liu, W.,Jia, Y., Sermanet, P.,Reed, S.,Anguel ov, D.,Erhan, D., Vanhoucke, V., and Rabinovich, A. Goingdeeper withconvolutions. CoRR,abs/1409.4842, 2014. Wei, Y., Xia, W., Huang, J., Ni, B., Dong, J., Zhao, Y., and Yan , S. CNN: Single-label to multi-label. CoRR, abs/1406.5726, 2014. Zeiler, M. D. and Fergus, R. Visualizing and understanding c onvolutional networks. CoRR, abs/1311.2901, 2013. PublishedinProc. ECCV,2014. A LOCALISATION In the main bodyof the paper we have consideredthe classifica tion task of the ILSVRC challenge,

1409.1556

Very Deep Convolutional Networks for Large-Scale Image Recognition

In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

http://arxiv.org/pdf/1409.1556

cs.CV

20140904

20150410

1409.1556

44

A LOCALISATION In the main bodyof the paper we have consideredthe classifica tion task of the ILSVRC challenge, and performed a thorough evaluation of ConvNet architectur es of different depth. In this section, we turn to the localisation task of the challenge, which we ha ve won in 2014 with 25.3%error. It can be seen as a special case of object detection, where a sing le object bounding box should be predictedforeach ofthe top-5classes, irrespectiveof the actual numberofobjectsof the class. For thiswe adoptthe approachof Sermanetet al. (2014), the winn ersof the ILSVRC-2013localisation challenge,withafewmodifications. Ourmethodisdescribed inSect.A.1andevaluatedinSect.A.2. A.1 L OCALISATION CONVNET To perform object localisation, we use a very deep ConvNet, w here the last fully connected layer predicts the bounding box location instead of the class scor es. A bounding box is represented by a 4-D vector storing its center coordinates, width, and heig ht. There is a choice of whether the

1409.1556

Very Deep Convolutional Networks for Large-Scale Image Recognition

In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

http://arxiv.org/pdf/1409.1556

cs.CV

20140904

20150410

1409.1556

45

a 4-D vector storing its center coordinates, width, and heig ht. There is a choice of whether the boundingbox prediction is shared across all classes (singl e-class regression, SCR (Sermanetet al., 2014))orisclass-specific(per-classregression,PCR).In theformercase,thelastlayeris4-D,while in the latter it is 4000-D (since there are 1000 classes in the dataset). Apart from the last bounding boxpredictionlayer,weuse theConvNetarchitectureD (Tab le1),whichcontains16weightlayers andwasfoundtobe thebest-performingin theclassification task (Sect.4). Training. Training of localisation ConvNets is similar to that of the c lassification ConvNets (Sect.3.1). Themaindifferenceisthatwereplacethelogis ticregressionobjectivewithaEuclidean loss,whichpenalisesthedeviationofthepredictedboundi ngboxparametersfromtheground-truth. We trainedtwo localisation models, each on a single scale: S= 256andS= 384(due to the time

1409.1556

Very Deep Convolutional Networks for Large-Scale Image Recognition

In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

http://arxiv.org/pdf/1409.1556

cs.CV

20140904

20150410

1409.1556

46

We trainedtwo localisation models, each on a single scale: S= 256andS= 384(due to the time constraints,we didnot use trainingscale jitteringforour ILSVRC-2014submission). Trainingwas initialised with the correspondingclassification models ( trained on the same scales), and the initial learning rate was set to 10−3. We exploredboth fine-tuningall layers and fine-tuningonly the first two fully-connected layers, as done in (Sermanetetal., 201 4). The last fully-connected layer was initialisedrandomlyandtrainedfromscratch. Testing. We consider two testing protocols. The first is used for compa ring different network modifications on the validation set, and considers only the b oundingbox prediction for the ground truth class (to factor out the classification errors). The bo unding box is obtained by applying the networkonlyto thecentralcropoftheimage. The second, fully-fledged, testing procedure is based on the dense application of the localisation ConvNet to the whole image, similarly to the classification t ask (Sect. 3.2). The difference is that

1409.1556

Very Deep Convolutional Networks for Large-Scale Image Recognition

In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

http://arxiv.org/pdf/1409.1556

cs.CV

20140904

20150410

1409.1556

47

ConvNet to the whole image, similarly to the classification t ask (Sect. 3.2). The difference is that instead of the class score map, the output of the last fully-c onnected layer is a set of bounding box predictions. To come up with the final prediction, we util ise the greedy merging procedure of Sermanetetal. (2014), which first merges spatially close predictions (by averaging their coordinates), and then rates them based on the class scores, obta ined from the classification ConvNet. When several localisation ConvNets are used, we first take th e union of their sets of boundingbox predictions, and then run the mergingprocedureon the union . We did not use the multiple pooling 10 Publishedasa conferencepaperat ICLR2015 offsets technique of Sermanetetal. (2014), which increase s the spatial resolution of the bounding boxpredictionsandcanfurtherimprovetheresults. A.2 L OCALISATION EXPERIMENTS

1409.1556

Very Deep Convolutional Networks for Large-Scale Image Recognition

In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

http://arxiv.org/pdf/1409.1556

cs.CV

20140904

20150410

1409.1556

48

boxpredictionsandcanfurtherimprovetheresults. A.2 L OCALISATION EXPERIMENTS In this section we first determine the best-performinglocal isation setting (using the first test protocol), and then evaluate it in a fully-fledged scenario (the se cond protocol). The localisation error is measured according to the ILSVRC criterion (Russakovsky et al., 2014), i.e. the bounding box predictionis deemed correctif its intersectionoverunion ratio with the ground-truthboundingbox isabove0.5. Settings comparison. As can be seen from Table 8, per-class regression (PCR) outpe rforms the class-agnostic single-class regression (SCR), which diff ers from the findings of Sermanetetal. (2014), where PCR was outperformed by SCR. We also note that fi ne-tuning all layers for the localisation task leads to noticeablybetter results than fine -tuningonly the fully-connectedlayers(as donein(Sermanetet al.,2014)). Intheseexperiments,thes mallestimagessidewassetto S= 384; theresultswith S= 256exhibitthesamebehaviourandarenotshownforbrevity.

1409.1556

Very Deep Convolutional Networks for Large-Scale Image Recognition

In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

http://arxiv.org/pdf/1409.1556

cs.CV

20140904

20150410

1409.1556

49

theresultswith S= 256exhibitthesamebehaviourandarenotshownforbrevity. Table 8:Localisation error for different modifications with the simplified testing protocol: the boundingbox is predictedfrom a single central image crop, a nd the ground-truthclass is used. All ConvNet layers (except for the last one) have the configurati on D (Table 1), while the last layer performseithersingle-classregression(SCR) orper-clas sregression(PCR). Fine-tunedlayers regression type GTclass localisationerror 1st and2nd FCSCR 36.4 PCR 34.3 all PCR 33.1 Fully-fledgedevaluation. Havingdeterminedthebestlocalisationsetting(PCR,fine- tuningofall layers),we nowapply it in the fully-fledgedscenario,where the top-5class labelsare predictedusing our best-performingclassification system (Sect. 4.5), and multiple densely-computedbounding

1409.1556

Very Deep Convolutional Networks for Large-Scale Image Recognition

In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

http://arxiv.org/pdf/1409.1556

cs.CV

20140904

20150410

1409.1556

50

box predictions are merged using the method of Sermanetetal . (2014). As can be seen from Table 9, applicationof the localisationConvNetto the whole i magesubstantiallyimprovesthe results compared to using a center crop (Table 8), despite using the t op-5 predicted class labels instead of thegroundtruth. Similarlytotheclassificationtask(Sect .4),testingatseveralscalesandcombining thepredictionsofmultiplenetworksfurtherimprovesthep erformance. Table9:Localisationerror smallestimage side top-5localisationerror (%) train(S) test(Q) val. test. 256 256 29.5 384 384 28.2 26.7 384 352,384 27.5 fusion: 256/256 and 384/352,384 26.9 25.3 Comparison with the state of the art. We compare our best localisation result with the state of the art in Table 10. With 25.3%test error, our “VGG” team won the localisation challenge of ILSVRC-2014 (Russakovskyet al., 2014). Notably, our resul ts are considerably better than those of the ILSVRC-2013winnerOverfeat(Sermanetet al., 2014), even thoughwe used less scales and

1409.1556

Very Deep Convolutional Networks for Large-Scale Image Recognition

In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

http://arxiv.org/pdf/1409.1556

cs.CV

20140904

20150410

1409.1556

51

of the ILSVRC-2013winnerOverfeat(Sermanetet al., 2014), even thoughwe used less scales and did not employ their resolution enhancement technique. We e nvisage that better localisation performance can be achieved if this technique is incorporated i nto our method. This indicates the performanceadvancementbroughtbyourverydeepConvNets– wegotbetterresultswithasimpler localisationmethod,buta morepowerfulrepresentation. B GENERALISATION OF VERYDEEPFEATURES In the previous sections we have discussed training and eval uation of very deep ConvNets on the ILSVRC dataset. In this section, we evaluate our ConvNets, p re-trained on ILSVRC, as feature 11 Publishedasa conferencepaperat ICLR2015 Table 10: Comparison with the state of the art in ILSVRC localisation . Our methodis denoted as“VGG”. Method top-5val. error (%) top-5 testerror (%) VGG 26.9 25.3 GoogLeNet (Szegedyet al., 2014) - 26.7 OverFeat (Sermanet etal.,2014) 30.0 29.9 Krizhevsky et al.(Krizhevsky et al.,2012) - 34.2

1409.1556

Very Deep Convolutional Networks for Large-Scale Image Recognition

In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

http://arxiv.org/pdf/1409.1556

cs.CV

20140904

20150410

1409.1556

52

Krizhevsky et al.(Krizhevsky et al.,2012) - 34.2 extractors on other, smaller, datasets, where training lar ge models from scratch is not feasible due to over-fitting. Recently, there has been a lot of interest in such a use case (Zeiler&Fergus, 2013; Donahueet al., 2013; Razavianet al., 2014; Chatfieldet al., 2014), as it turns out that deep image representations,learntonILSVRC,generalisewelltoothe rdatasets,wheretheyhaveoutperformed hand-crafted representations by a large margin. Following that line of work, we investigate if our modelsleadtobetterperformancethanmoreshallowmodelsu tilisedinthestate-of-the-artmethods. In this evaluation, we consider two models with the best clas sification performance on ILSVRC (Sect.4)–configurations“Net-D”and“Net-E”(whichwemade publiclyavailable). To utilise the ConvNets, pre-trained on ILSVRC, for image cl assification on other datasets, we

1409.1556

Very Deep Convolutional Networks for Large-Scale Image Recognition

In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

http://arxiv.org/pdf/1409.1556

cs.CV

20140904

20150410

1409.1556

53

To utilise the ConvNets, pre-trained on ILSVRC, for image cl assification on other datasets, we remove the last fully-connected layer (which performs 1000 -way ILSVRC classification), and use 4096-Dactivationsofthepenultimatelayerasimagefeatur es,whichareaggregatedacrossmultiple locations and scales. The resulting image descriptor is L2-normalised and combined with a linear SVM classifier, trained on the target dataset. For simplicit y, pre-trained ConvNet weights are kept fixed(nofine-tuningisperformed). Aggregation of features is carried out in a similar manner to our ILSVRC evaluation procedure (Sect. 3.2). Namely, an image is first rescaled so that its sma llest side equals Q, and then the network is densely applied over the image plane (which is possib le when all weight layers are treated as convolutional). We then perform global average pooling o n the resulting feature map, which producesa 4096-Dimage descriptor. The descriptor is then a veraged with the descriptor of a horizontally flipped image. As was shown in Sect. 4.2, evaluation over multiple scales is beneficial, so

1409.1556

Very Deep Convolutional Networks for Large-Scale Image Recognition

In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

http://arxiv.org/pdf/1409.1556

cs.CV

20140904

20150410