Recurrent Neural Network Training with Dark Knowledge Transfer

TL;DR

This paper demonstrates that using a deep neural network as a teacher can effectively transfer knowledge to train recurrent neural networks, specifically LSTMs, in speech recognition tasks with limited data.

Contribution

It introduces a novel approach where a weaker DNN model is used as a teacher to train RNNs, showing success without special tricks in limited data scenarios.

Findings

RNNs trained with DNN teachers perform well in ASR tasks.

Knowledge transfer from weaker to stronger models is effective.

Training succeeds without additional tricks even with limited data.

Abstract

Recurrent neural networks (RNNs), particularly long short-term memory (LSTM), have gained much attention in automatic speech recognition (ASR). Although some successful stories have been reported, training RNNs remains highly challenging, especially with limited training data. Recent research found that a well-trained model can be used as a teacher to train other child models, by using the predictions generated by the teacher model as supervision. This knowledge transfer learning has been employed to train simple neural nets with a complex one, so that the final performance can reach a level that is infeasible to obtain by regular training. In this paper, we employ the knowledge transfer learning approach to train RNNs (precisely LSTM) using a deep neural network (DNN) model as the teacher. This is different from most of the existing research on knowledge transfer learning, since the teacher (DNN) is assumed to be weaker than the child (RNN); however, our experiments on an ASR task showed that it works fairly well: without applying any tricks on the learning scheme, this approach can train RNNs successfully even with limited training data.