Compression of Deep Neural Networks on the Fly

TL;DR

This paper presents a novel method for compressing deep neural networks during training by adding regularization and combining it with product quantization, enabling significant size reduction suitable for resource-limited devices.

Contribution

The authors introduce a new on-the-fly compression technique that integrates regularization into training and combines it with product quantization for improved compression rates.

Findings

Achieves larger compression rates than existing methods on MNIST and CIFAR10.

Compresses neural networks during training, reducing storage without significant accuracy loss.

Combines regularization and product quantization for effective model size reduction.

Abstract

Thanks to their state-of-the-art performance, deep neural networks are increasingly used for object recognition. To achieve these results, they use millions of parameters to be trained. However, when targeting embedded applications the size of these models becomes problematic. As a consequence, their usage on smartphones or other resource limited devices is prohibited. In this paper we introduce a novel compression method for deep neural networks that is performed during the learning phase. It consists in adding an extra regularization term to the cost function of fully-connected layers. We combine this method with Product Quantization (PQ) of the trained weights for higher savings in storage consumption. We evaluate our method on two data sets (MNIST and CIFAR10), on which we achieve significantly larger compression rates than state-of-the-art methods.