Wojciech Zaremba, Ilya Sutskever, Oriol Vinyals
We present a simple regularization technique for Recurrent Neural Networks (RNNs) with Long Short-Term Memory (LSTM) units. Dropout, the most successful technique for regularizing neural networks, does not work well with RNNs and LSTMs. In this paper, we show how to correctly apply dropout to LSTMs, and show that it substantially reduces overfitting on a variety of tasks. These tasks include language modeling, speech recognition, image caption generation, and machine translation.
| Task | Dataset | Metric | Value | Model |
|---|---|---|---|---|
| Machine Translation | WMT2014 English-French | BLEU score | 29.03 | Regularized LSTM |
| Language Modelling | Penn Treebank (Word Level) | Test perplexity | 78.4 | Zaremba et al. (2014) - LSTM (large) |
| Language Modelling | Penn Treebank (Word Level) | Validation perplexity | 82.2 | Zaremba et al. (2014) - LSTM (large) |
| Language Modelling | Penn Treebank (Word Level) | Test perplexity | 82.7 | Zaremba et al. (2014) - LSTM (medium) |
| Language Modelling | Penn Treebank (Word Level) | Validation perplexity | 86.2 | Zaremba et al. (2014) - LSTM (medium) |