Learning Compact Boolean Networks

Shengpu Wang, Yuhao Mao, Yani Zhang, Martin Vechev · 2026

Read Learning Compact Boolean Networks by Shengpu Wang, Yuhao Mao, Yani Zhang, Martin Vechev (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-05830.txt.

Description

Neural networks built only from Boolean operations promise nanosecond inference on cheap hardware, but learning compact and accurate ones is hard because their structure is discrete and combinatorial. Shengpu Wang, Yuhao Mao, Yani Zhang, and Martin Vechev attack the problem with three complementary ideas. The result is a learning method for Boolean networks aimed at resource constrained deployments.

First, a parameter free strategy learns effective connections. Second, a compact convolutional Boolean architecture exploits spatial locality with fewer operations than existing kernels. Third, an adaptive discretization procedure shrinks the accuracy loss when a relaxed network is converted into a discrete one. On standard vision benchmarks the method improved the accuracy efficiency frontier, reaching higher accuracy with far fewer Boolean operations. On dedicated hardware, their MNIST model hit high accuracy with single digit nanosecond latency and a much smaller circuit than the prior best.

Catalog record

CREATORShengpu Wang, Yuhao Mao, Yani Zhang, Martin Vechev
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE50 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2602-05830

Browse the full library · About the Tower of Babel