Training Quantized Deep Neural Networks via Cooperative Coevolution

TL;DR

This paper introduces a novel evolutionary algorithm-based method for training quantized deep neural networks without full-precision operations, achieving comparable accuracy to full-precision models.

Contribution

It is the first to apply evolutionary algorithms with cooperative coevolution to train quantized DNNs, addressing a key limitation of existing methods.

Findings

Surpasses previous quantization methods in accuracy

Successfully trains 4-bit ResNet-20 with full-precision accuracy on Cifar-10

Demonstrates effectiveness of cooperative coevolution in large-scale discrete optimization

Abstract

This work considers a challenging Deep Neural Network(DNN) quantization task that seeks to train quantized DNNs without involving any full-precision operations. Most previous quantization approaches are not applicable to this task since they rely on full-precision gradients to update network weights. To fill this gap, in this work we advocate using Evolutionary Algorithms (EAs) to search for the optimal low-bits weights of DNNs. To efficiently solve the induced large-scale discrete problem, we propose a novel EA based on cooperative coevolution that repeatedly groups the network weights based on the confidence in their values and focuses on optimizing the ones with the least confidence. To the best of our knowledge, this is the first work that applies EAs to train quantized DNNs. Experiments show that our approach surpasses previous quantization approaches and can train a 4-bit ResNet-20 on the Cifar-10 dataset with the same test accuracy as its full-precision counterpart.