Compressing Deep Neural Networks Using Explainable AI

TL;DR

This paper introduces a novel DNN compression method leveraging explainable AI, specifically Layer-wise Relevance Propagation, to prune and quantize models efficiently, achieving significant size reduction with minimal accuracy loss.

Contribution

The paper presents a new XAI-based compression technique that uses relevance scores for pruning and mixed-precision quantization, outperforming existing methods in size reduction and accuracy.

Findings

Reduces model size by 64%

Improves accuracy by 42% over previous XAI-based methods

Effective pruning and quantization guided by relevance scores

Abstract

Deep neural networks (DNNs) have demonstrated remarkable performance in many tasks but it often comes at a high computational cost and memory usage. Compression techniques, such as pruning and quantization, are applied to reduce the memory footprint of DNNs and make it possible to accommodate them on resource-constrained edge devices. Recently, explainable artificial intelligence (XAI) methods have been introduced with the purpose of understanding and explaining AI methods. XAI can be utilized to get to know the inner functioning of DNNs, such as the importance of different neurons and features in the overall performance of DNNs. In this paper, a novel DNN compression approach using XAI is proposed to efficiently reduce the DNN model size with negligible accuracy loss. In the proposed approach, the importance score of DNN parameters (i.e. weights) are computed using a gradient-based XAI technique called Layer-wise Relevance Propagation (LRP). Then, the scores are used to compress the DNN as follows: 1) the parameters with the negative or zero importance scores are pruned and removed from the model, 2) mixed-precision quantization is applied to quantize the weights with higher/lower score with higher/lower number of bits. The experimental results show that, the proposed compression approach reduces the model size by 64% while the accuracy is improved by 42% compared to the state-of-the-art XAI-based compression method.