MariusAnje / SEU_NNFailure

Repo for paper "When Single Event Upset Meets Deep Neural Networks: Observations, Explorations, and Remedies"; Readme under construction, code not useful right now

SEU_NNFailure

This project tests Neural Networks' robustness against Single Event Upset.

Workflow

A pretrained Neural Networks, invert one of its weights' first bit, test the disrupted network on a validation set, then record the accuracy loss. Test every bits (or randomly sample some bits) of this network, analyze the effect of their inversion.

Experiments

Experiments are conducted mostly on XNOR-Net

XNOR-Net ImpeImplementation

Used an PyTorch implementation of the XNOR-Net. Major networks are as follows:

Dataset	Network	Accuracy
MNIST	LeNet-5	99.23%
CIFAR-10	Network-in-Network (NIN)	86.28%
ImageNet	AlexNet	Top-1: 44.87% Top-5: 69.70%

MNIST

CIFAR-10

ImageNet