Emulating the Effects of Radiation-Induced Soft-Errors for the Reliability Assessment of Neural Networks

Abstract

Convolutional Neural Networks (CNNs) are currently one of the most widely used predictive models in machine learning. Recent studies have demonstrated that hardware faults induced by radiation fields, including cosmic rays, may significantly impact the CNN inference leading to wrong predictions. Therefore, ensuring the reliability of CNNs is crucial, especially for safety-critical systems. In the literature, several works propose reliability assessments of CNNs mainly based on statistically injected faults. This work presents a software emulator capable of injecting real faults retrieved from radiation tests. Specifically, from the device characterisation of a DRAM memory, we extracted event rates and fault models. The software emulator can reproduce their incidence and access their effect on CNN applications with a reliability assessment precision close to the physical one. Radiation-based physical injections and emulator-based injections are performed on three CNNs (LeNet-5) exploiting different data representations. Their outcomes are compared, and the software results evidence that the emulator is able to reproduce the faulty behaviours observed during the radiation tests for the targeted CNNs. This approach leads to a more concise use of radiation experiments since the extracted fault models can be reused to explore different scenarios (e.g., impact on a different application).

Testability and Dependability of AI Hardware: Survey, Trends, Challenges, and Perspectives

“…In [69], a fault injection framework is proposed that reproduces fault models and event rates extracted from radiation tests. The ultimate goal is to have the flexibility of a softwarebased fault injector with a reliability assessment precision close to this of an accelerated neutron beam radiation-based fault injection experiment in a realistic harsh environment.…”

A Survey on Deep Learning Resilience Assessment Methodologies

“…Furthermore, the authors in [16] studied the impact of neutron irradiation on the Hy-perRAM memory, which stored the weights of the CNN-based application. In this way, the source of error was isolated and the focus was on the CNN weights.…”