A Controlled Diagnostic Study of Hardware-Induced Distortions in Hardware-Aware Training

2026-05-10

AI Summary

This work addresses the challenge that hardware-aware training often fails to universally compensate for diverse hardware non-idealities, necessitating a clear delineation of the boundaries within which trainable mitigation is feasible. The authors propose a diagnostic framework that models hardware non-idealities as structured perturbations to forward operators and systematically evaluates their compatibility with gradient-based optimization through three theoretical criteria: expected gradient consistency, bounded gradient variance, and non-degenerate sensitivity. By integrating forward perturbation modeling, rigorous gradient analysis, and controlled experiments across six representative distortion types—including read noise and IR drop—the study rigorously distinguishes between compensable and inherently uncompensable hardware distortions. These findings provide both theoretical grounding and practical guidance for hardware-software co-design, highlighting that certain non-idealities fundamentally require correction at the circuit or architectural level rather than through algorithmic adaptation alone.

Abstract

Hardware-aware training (HAT) is widely used to improve the robustness of neural networks on non-ideal AI accelerators, such as analog in-memory computing (IMC) systems. However, not all hardware-induced distortions are equally compensable by training. This paper presents a diagnostic framework that models hardware non-idealities as structured perturbations of the forward operator and evaluates their compatibility with gradient-based optimization. We analyze six representative perturbation classes--read noise, variability, drift, stuck-at faults, IR-drop, and ADC discretization--and identify three key diagnostics: gradient expectation consistency, bounded gradient variance, and non-degenerate sensitivity. Our results show a clear separation between perturbations that can be compensated by HAT and those that consistently break optimization. This provides practical guidance for hardware-software co-design, clarifying which non-idealities can be addressed at the training level and which require circuit-, architecture-, or calibration-level mitigation. This study should be interpreted as a controlled empirical analysis under vanilla forward-perturbation HAT, rather than as a universal theory of hardware-aware training.

Problem

• hardware-aware training
• hardware-induced distortions
• non-idealities
• gradient-based optimization
• AI accelerators

Innovation

• hardware-aware training
• structured perturbations
• gradient diagnostics
• non-idealities
• co-design