On Fast Dropout and its Applicability to Recurrent Networks

Abstract

Recurrent Neural Networks (RNNs) are rich models for the processing of sequential data. Recent work on advancing the state of the art has been focused on the optimization or modelling of RNNs, mostly motivated by adressing the problems of the vanishing and exploding gradients. The control of overfitting has seen considerably less attention. This paper contributes to that by analyzing fast dropout, a recent regularization method for generalized linear models and neural networks from a back-propagation inspired perspective. We show that fast dropout implements a quadratic form of an adaptive, per-parameter regularizer, which rewards large weights in the light of underfitting, penalizes them for overconfident predictions and vanishes at minima of an unregularized training loss. The derivatives of that regularizer are exclusively based on the training error signal. One consequence of this is the absense of a global weight attractor, which is particularly appealing for RNNs, since the dynamics are not biased towards a certain regime. We positively test the hypothesis that this improves the performance of RNNs on four musical data sets.

The experiments for the Penn Treebank corpus were erroneous and have been stripped from this version

References in corpus (3)

• Improving neural networks by preventing co-adaptation of feature detectors
• How to Construct Deep Recurrent Neural Networks
• Speech Recognition with Deep Recurrent Neural Networks

Cited by in corpus (16)

• Deep Learning in Neural Networks: An Overview
• Recurrent Neural Network Regularization
• A Theoretically Grounded Application of Dropout in Recurrent Neural Networks
• How to Construct Deep Recurrent Neural Networks
• Learning Stochastic Recurrent Networks
• A Deep Structured Model with Radius-Margin Bound for 3D Human Activity Recognition
• DeepCare: A Deep Dynamic Memory Model for Predictive Medicine
• Progressive Memory Banks for Incremental Domain Adaptation
• Bayesian Sparsification of Recurrent Neural Networks
• Pushing the bounds of dropout
• Identifying Audio Adversarial Examples via Anomalous Pattern Detection
• Scalable Bayesian Learning of Recurrent Neural Networks for Language Modeling
• Learned-Norm Pooling for Deep Feedforward and Recurrent Neural Networks
• Variational inference of latent state sequences using Recurrent Networks
• Regularizing Recurrent Networks - On Injected Noise and Norm-based Methods
• An Analysis of Dropout for Matrix Factorization