BOLD: Boolean Logic Deep Learning · HackerPapers discussions

BOLD: Boolean Logic Deep Learning — mentioned by cubefox

Mon, 20 Jul 2026 15:33:18 GMT

There was also a NeurIPS paper which successfully trained 1-bit models directly without floating point weights or any quantization: https://arxiv.org/abs/2405.16339 I didn't see any further work in this direction however. The paper seems to have been mostly overlooked.

Paper and discussion history. A mention is not an endorsement.

BOLD: Boolean Logic Deep Learning — mentioned by cubefox

Thu, 10 Apr 2025 12:43:58 GMT

Much lower indeed. Even Boolean functions (e.g. AND) are differentiable (though not exactly in the Newton/Leibniz sense) which can be used for backpropagation. They allow for an optimizer similar to stochastic gradient descent. There is a paper on it: https://arxiv.org/abs/2405.16339 It seems to me that floating point math (matrix multiplication) will over time mostly disappear from ML chips, as Boolean operations are much faster both in training an inference. But currently they are still optimized for FP rather than Boolean operations.

Paper and discussion history. A mention is not an endorsement.

BOLD: Boolean Logic Deep Learning — HN submission

Mon, 24 Mar 2025 14:08:03 GMT

Bold: Boolean Logic Deep Learning

2 points; 1 discussion reply on HN. Paper and source history. A submission is not an endorsement.