Boolean neural networks offer hardware-efficient alternatives to real-valued models. While quantization is common, purely Boolean training remains underexplored. We present a practical method for purely Boolean backpropagation for networks based on a single specific gate we chose, operating directly in Boolean algebra involving no numerics. Initial experiments confirm its feasibility.
Cite
@article{arxiv.2505.03791,
title = {Practical Boolean Backpropagation},
author = {Simon Golbert},
journal= {arXiv preprint arXiv:2505.03791},
year = {2025}
}