AIDX: Adaptive Inference Scheme to Mitigate State-Drift in Memristive VMM Accelerators

TL;DR

AIDX is an adaptive inference scheme that dynamically adjusts input pulses to reduce memristance drift effects, significantly improving neural network accuracy and image reconstruction quality in memristive VMM accelerators.

Contribution

The paper introduces a novel adaptive inference method that mitigates memristance drift in memristive accelerators through optimized pulse adjustments, enhancing neural network performance.

Findings

60% improvement in CNN accuracy on CIFAR10 after 10,000 inferences

78.6% reduction in error for image reconstruction

Effective mitigation of memristance drift effects

Abstract

An adaptive inference method for crossbar (AIDX) is presented based on an optimization scheme for adjusting the duration and amplitude of input voltage pulses. AIDX minimizes the long-term effects of memristance drift on artificial neural network accuracy. The sub-threshold behavior of memristor has been modeled and verified by comparing with fabricated device data. The proposed method has been evaluated by testing on different network structures and applications, e.g., image reconstruction and classification tasks. The results showed an average of 60% improvement in convolutional neural network (CNN) performance on CIFAR10 dataset after 10000 inference operations as well as 78.6% error reduction in image reconstruction.