Improving deep neural networks for LVCSR using rectified linear units and dropout

TLDR

Deep neural networks have surpassed GMMs in large‑vocabulary speech recognition, and techniques such as dropout and ReLU units—known to improve generalization and training speed in vision tasks—are being explored to enhance acoustic models. On a 50‑hour English Broadcast News task, ReLU‑based DNNs trained with dropout achieved a 4.2 % relative gain over sigmoid DNNs and a 14.4 % gain over a strong GMM/HMM baseline, with minimal hyper‑parameter tuning.

Abstract

Recently, pre-trained deep neural networks (DNNs) have outperformed traditional acoustic models based on Gaussian mixture models (GMMs) on a variety of large vocabulary speech recognition benchmarks. Deep neural nets have also achieved excellent results on various computer vision tasks using a random "dropout" procedure that drastically improves generalization error by randomly omitting a fraction of the hidden units in all layers. Since dropout helps avoid over-fitting, it has also been successful on a small-scale phone recognition task using larger neural nets. However, training deep neural net acoustic models for large vocabulary speech recognition takes a very long time and dropout is likely to only increase training time. Neural networks with rectified linear unit (ReLU) non-linearities have been highly successful for computer vision tasks and proved faster to train than standard sigmoid units, sometimes also improving discriminative performance. In this work, we show on a 50-hour English Broadcast News task that modified deep neural networks using ReLUs trained with dropout during frame level training provide an 4.2% relative improvement over a DNN trained with sigmoid units, and a 14.4% relative improvement over a strong GMM/HMM system. We were able to obtain our results with minimal human hyper-parameter tuning using publicly available Bayesian optimization code.