Building on work of Charton, we train small transformer models to calculate the M\"{o}bius function μ(n) and the squarefree indicator function μ2(n). The models attain nontrivial predictive power. We apply a mixture of additional models and feature scoring to give a theoretical explanation.
Cite
@article{arxiv.2502.10335,
title = {Studying number theory with deep learning: a case study with the M\"obius and squarefree indicator functions},
author = {David Lowry-Duda},
journal= {arXiv preprint arXiv:2502.10335},
year = {2025}
}