ECQ$^{\text{x}}$: Explainability-Driven Quantization for Low-Bit and Sparse DNNs

TL;DR

ECQ$^{ ext{x}}$ introduces a novel explainability-driven quantization method that produces ultra low-bit and sparse neural networks, significantly reducing model size while maintaining or improving performance across various datasets.

Contribution

The paper proposes a new quantization approach that incorporates explainability and information theory to preserve important weights, enabling ultra low-bit and sparse DNNs with high compression.

Findings

Achieves 2-5 bit quantization with maintained or improved accuracy.

Up to 103× reduction in model size compared to full-precision models.

Effective across multiple datasets and model types.

Abstract

The remarkable success of deep neural networks (DNNs) in various applications is accompanied by a significant increase in network parameters and arithmetic operations. Such increases in memory and computational demands make deep learning prohibitive for resource-constrained hardware platforms such as mobile devices. Recent efforts aim to reduce these overheads, while preserving model performance as much as possible, and include parameter reduction techniques, parameter quantization, and lossless compression techniques. In this chapter, we develop and describe a novel quantization paradigm for DNNs: Our method leverages concepts of explainable AI (XAI) and concepts of information theory: Instead of assigning weight values based on their distances to the quantization clusters, the assignment function additionally considers weight relevances obtained from Layer-wise Relevance Propagation (LRP) and the information content of the clusters (entropy optimization). The ultimate goal is to preserve the most relevant weights in quantization clusters of highest information content. Experimental results show that this novel Entropy-Constrained and XAI-adjusted Quantization (ECQ$^{\text{x}}$) method generates ultra low-precision (2-5 bit) and simultaneously sparse neural networks while maintaining or even improving model performance. Due to reduced parameter precision and high number of zero-elements, the rendered networks are highly compressible in terms of file size, up to $103\times$ compared to the full-precision unquantized DNN model. Our approach was evaluated on different types of models and datasets (including Google Speech Commands, CIFAR-10 and Pascal VOC) and compared with previous work.