Scalable Bayesian learning of recurrent neural networks for language modeling


Conference Paper

© 2017 Association for Computational Linguistics. Recurrent neural networks (RNNs) have shown promising performance for language modeling. However, traditional training of RNNs using back-propagation through time often suffers from overfitting. One reason for this is that stochastic optimization (used for large training sets) does not provide good estimates of model uncertainty. This paper leverages recent advances in stochastic gradient Markov Chain Monte Carlo (also appropriate for large training sets) to learn weight uncertainty in RNNs. It yields a principled Bayesian learning algorithm, adding gradient noise during training (enhancing exploration of the model-parameter space) and model averaging when testing. Extensive experiments on various RNN models and across a broad range of applications demonstrate the superiority of the proposed approach relative to stochastic optimization.

Full Text

Duke Authors

Cited Authors

  • Gan, Z; Li, C; Chen, C; Pu, Y; Su, Q; Carin, L

Published Date

  • January 1, 2017

Published In

  • Acl 2017 55th Annual Meeting of the Association for Computational Linguistics, Proceedings of the Conference (Long Papers)

Volume / Issue

  • 1 /

Start / End Page

  • 321 - 331

International Standard Book Number 13 (ISBN-13)

  • 9781945626753

Digital Object Identifier (DOI)

  • 10.18653/v1/P17-1030

Citation Source

  • Scopus