works on
algebraic representability 1grokking 1holomorphic activation 1modular arithmetic 1neural network capacity 1

From the 1 of 11 linked papers with an AI index.

collaborators
Showing cs.LGShow all

1 paper · 1 filter

cs.LG2026

Algebraic Representability as the Limiting Regime of Grokking: An Exactly Solvable Model with Holomorphic Activations

Chon-Fai Kam, Xavier Cadet, Miloud Bessafi +1

The paper analyzes two‑layer neural networks with holomorphic monomial activations (σ(z)=z^k) on modular arithmetic tasks, providing an exact algebraic condition for when a target…