Reproduce “Scalable fault vulnerability analysis for ReRAM-based process-in-memory CNN accelerators” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Aniseh Dorostkar, Hamed Farbeh, Hamid R. Zarandi, Meysam Sharifzadeh

The authors estimate the impact of ReRAM stuck-at faults on CNN weights using loss gradients and curvature, aggregating sensitivities into weight, kernel, and layer vulnerability rankings. Selected VGG19 kernels show worse injected-fault outcomes at higher scores. ResNet18 rank correlations range from 0.66 to 0.69, while SimpleNet’s global correlations are weak. SimpleNet analysis takes 4.696 seconds versus an estimated 4.9 days.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.