Reproduce “BitLogic: Training Framework for Gradient-Based FPGA-Native Neural Networks” 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

Simon Bührer, Andreas Plesner, Aczel Till, Roger Wattenhofer

BitLogic makes Boolean lookup networks configurable across five choices: input encoding, connectivity, fan-in, node parameterization, and output head, enabling controlled comparisons under one training protocol. Its selected LightLUT-based configuration outperformed retrained priors in 62 of 62 comparable cells, including 58.06% versus 53.02% on CIFAR-10 at width 64,000. Results are limited to two-layer image classification.

Summary by alphaXiv.

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