Mind the Gap: Removing the Discretization Gap in Differentiable Logic Gate Networks
Shakir Yousefi, Andreas Plesner, Till Aczel, Roger Wattenhofer · 2025
Read Mind the Gap: Removing the Discretization Gap in Differentiable Logic Gate Networks by Shakir Yousefi, Andreas Plesner, Till Aczel, Roger Wattenhofer (2025) 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-2506-07500.txt.
Description
Shakir Yousefi, Andreas Plesner, Till Aczel, and Roger Wattenhofer attack a training flaw in logic gate networks. Their 2025 paper concerns networks built from logic gates, prized for cheap, low-energy classification. Training them is slow, and nearly half the gates often end up unused, so the deployed circuit falls short of its trained promise, a gap the authors name.
Their fix injects a special noise during training while passing gradients through cleanly. Training ran several times faster, the gap between training and final accuracy nearly vanished, and unused gates disappeared entirely. The authors trace the effect to an implicit smoothing of the loss surface. Efficient gate networks could bring capable image classification to tiny devices that cannot afford standard neural networks.
Catalog record
| CREATOR | Shakir Yousefi, Andreas Plesner, Till Aczel, Roger Wattenhofer |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 49 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2506-07500