Compression-aware Training of Deep Networks

TL;DR

This paper introduces a regularizer that encourages low-rank parameters during training, enabling the development of more compact neural networks that maintain performance, thus integrating compression considerations into the training process.

Contribution

It proposes a novel regularizer for training neural networks that explicitly incorporates compression, leading to more efficient models without sacrificing accuracy.

Findings

Learned models are significantly more compact.

Models maintain effectiveness comparable to state-of-the-art methods.

Training with the regularizer improves compression efficiency.

Abstract

In recent years, great progress has been made in a variety of application domains thanks to the development of increasingly deeper neural networks. Unfortunately, the huge number of units of these networks makes them expensive both computationally and memory-wise. To overcome this, exploiting the fact that deep networks are over-parametrized, several compression strategies have been proposed. These methods, however, typically start from a network that has been trained in a standard manner, without considering such a future compression. In this paper, we propose to explicitly account for compression in the training process. To this end, we introduce a regularizer that encourages the parameter matrix of each layer to have low rank during training. We show that accounting for compression during training allows us to learn much more compact, yet at least as effective, models than state-of-the-art compression techniques.