Sensitivity-Aware Mixed-Precision Quantization for ReRAM-based Computing-in-Memory

TL;DR

This paper introduces a sensitivity-aware mixed-precision quantization method tailored for ReRAM-based Compute-In-Memory systems, significantly improving efficiency and power consumption while maintaining high neural network accuracy.

Contribution

It proposes a novel structured quantization approach that combines sensitivity analysis with mixed-precision strategies specifically for ReRAM CIM architectures.

Findings

Achieves 86.33% accuracy at 70% compression

Reduces power consumption by 40%

Enhances ReRAM Crossbar utilization

Abstract

Compute-In-Memory (CIM) systems, particularly those utilizing ReRAM and memristive technologies, offer a promising path toward energy-efficient neural network computation. However, conventional quantization and compression techniques often fail to fully optimize performance and efficiency in these architectures. In this work, we present a structured quantization method that combines sensitivity analysis with mixed-precision strategies to enhance weight storage and computational performance on ReRAM-based CIM systems. Our approach improves ReRAM Crossbar utilization, significantly reducing power consumption, latency, and computational load, while maintaining high accuracy. Experimental results show 86.33% accuracy at 70% compression, alongside a 40% reduction in power consumption, demonstrating the method's effectiveness for power-constrained applications.