@incollection{NIPS2014_5346,
title = {Sequence to Sequence Learning with Neural Networks},
author = {Sutskever, Ilya and Vinyals, Oriol and Le, Quoc V},
booktitle = {Advances in Neural Information Processing Systems 27},
editor = {Z. Ghahramani and M. Welling and C. Cortes and N. D. Lawrence and K. Q. Weinberger},
pages = {3104--3112},
year = {2014},
publisher = {Curran Associates, Inc.},
url = {http://papers.nips.cc/paper/5346-sequence-to-sequence-learning-with-neural-networks.pdf}
}
[link]
1. What is it?
They proposed an efficient way to learn sequence to sequence (seq2seq) model into the machine translation task.
2. What is amazing compared to previous studies?
They used 2 LSTM units for encoder and decoder.
3. Where is the key to technologies and techniques?
They want to calculate this probability, generating target sentence y from source sentence x.
To calculate this probability, they used seq2seq models (this model is also Reccurent Neural Network; RNN).
v is a hidden state of source sentence x in last state of encoder LSTM.
Hidden state h and estimated word yt are calculated as below;
To train this model, objective function is below;
T is target sentences and S is source sentences, this set is called parallel corpus.
a1da4
commented
4 years ago
0. Paper
@incollection{NIPS2014_5346, title = {Sequence to Sequence Learning with Neural Networks}, author = {Sutskever, Ilya and Vinyals, Oriol and Le, Quoc V}, booktitle = {Advances in Neural Information Processing Systems 27}, editor = {Z. Ghahramani and M. Welling and C. Cortes and N. D. Lawrence and K. Q. Weinberger}, pages = {3104--3112}, year = {2014}, publisher = {Curran Associates, Inc.}, url = {http://papers.nips.cc/paper/5346-sequence-to-sequence-learning-with-neural-networks.pdf} } [link]
1. What is it?
They proposed an efficient way to learn sequence to sequence (seq2seq) model into the machine translation task.
2. What is amazing compared to previous studies?
They used 2 LSTM units for encoder and decoder.
3. Where is the key to technologies and techniques?
To train this model, objective function is below;
T is target sentences and S is source sentences, this set is called parallel corpus.
4. How did validate it?
WMT14 English to French dataset.
5. Is there a discussion?
6. Which paper should read next?
Seq2seq with Attention