Controlling Information Capacity of Binary Neural Network

TL;DR

This paper introduces a Shannon entropy-based penalty method to train binary neural networks, maintaining their information capacity and significantly improving accuracy on multiple datasets.

Contribution

It presents a novel training approach that stabilizes information capacity in binary networks, reducing accuracy degradation.

Findings

Significant accuracy improvements on SVHN, CIFAR, and ImageNet datasets.

Maintains stable information capacity during training.

Outperforms existing binary network training methods.

Abstract

Despite the growing popularity of deep learning technologies, high memory requirements and power consumption are essentially limiting their application in mobile and IoT areas. While binary convolutional networks can alleviate these problems, the limited bitwidth of weights is often leading to significant degradation of prediction accuracy. In this paper, we present a method for training binary networks that maintains a stable predefined level of their information capacity throughout the training process by applying Shannon entropy based penalty to convolutional filters. The results of experiments conducted on SVHN, CIFAR and ImageNet datasets demonstrate that the proposed approach can statistically significantly improve the accuracy of binary networks.