Recurrent Neural Network Regularization

8 Sep 2014  ·  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.

PDF Abstract
Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Language Modelling Penn Treebank (Word Level) Zaremba et al. (2014) - LSTM (medium) Validation perplexity 86.2 # 30
Test perplexity 82.7 # 38
Language Modelling Penn Treebank (Word Level) Zaremba et al. (2014) - LSTM (large) Validation perplexity 82.2 # 29
Test perplexity 78.4 # 35
Machine Translation WMT2014 English-French Regularized LSTM BLEU score 29.03 # 46

Methods