Weight Evolution: Improving Deep Neural Networks Training through Evolving Inferior Weight Values

TL;DR

This paper introduces weight evolution, a novel method inspired by genetic algorithms, to improve neural network training by selectively reactivating and updating unimportant weights through crossover with important weights, enhancing performance especially in lightweight networks.

Contribution

The paper proposes a new weight reactivation method called weight evolution that combines global and local selection with crossover strategies, outperforming existing methods in neural network training.

Findings

WE outperforms existing reactivation methods.

WE improves lightweight network performance.

The method is compatible with various architectures.

Abstract

To obtain good performance, convolutional neural networks are usually over-parameterized. This phenomenon has stimulated two interesting topics: pruning the unimportant weights for compression and reactivating the unimportant weights to make full use of network capability. However, current weight reactivation methods usually reactivate the entire filters, which may not be precise enough. Looking back in history, the prosperity of filter pruning is mainly due to its friendliness to hardware implementation, but pruning at a finer structure level, i.e., weight elements, usually leads to better network performance. We study the problem of weight element reactivation in this paper. Motivated by evolution, we select the unimportant filters and update their unimportant elements by combining them with the important elements of important filters, just like gene crossover to produce better offspring, and the proposed method is called weight evolution (WE). WE is mainly composed of four strategies. We propose a global selection strategy and a local selection strategy and combine them to locate the unimportant filters. A forward matching strategy is proposed to find the matched important filters and a crossover strategy is proposed to utilize the important elements of the important filters for updating unimportant filters. WE is plug-in to existing network architectures. Comprehensive experiments show that WE outperforms the other reactivation methods and plug-in training methods with typical convolutional neural networks, especially lightweight networks. Our code is available at https://github.com/BZQLin/Weight-evolution.