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.
View on arXiv@article{golbert2025_2505.03791,
title={ Practical Boolean Backpropagation },
author={ Simon Golbert },
journal={arXiv preprint arXiv:2505.03791},
year={ 2025 }
}