🤖 AI Summary
This work addresses deterministic black-box polynomial identity testing (PIT) for depth-4 arithmetic circuits of the form $Σ^{[3]}ΠΣΠ^{[δ]}$ over arbitrary fields, under the restriction that one top-layer addition gate computes a square-free polynomial. For this restricted yet important circuit class, we present the first polynomial-time deterministic black-box PIT algorithm valid over fields of arbitrary characteristic. Our key technical advance lies in replacing classical geometric approaches—reliant on Sylvester–Gallai-type theorems—with tools from algebraic geometry: specifically, normalization of algebraic varieties and computation of integral closures of coordinate rings. This eliminates dependence on field characteristic. By integrating square-free decomposition with structural analysis of coordinate rings, we construct decidable algebraic invariants. This is the first deterministic black-box PIT algorithm for bounded-depth, layered depth-4 circuits that works uniformly across all characteristics, establishing a new algebraic foundation for PIT of bounded-depth arithmetic circuits.
📝 Abstract
In this paper, we initiate the study of deterministic PIT for $Sigma^{[k]}PiSigmaPi^{[delta]}$ circuits over fields of any characteristic, where $k$ and $delta$ are bounded. Our main result is a deterministic polynomial-time black-box PIT algorithm for $Sigma^{[3]}PiSigmaPi^{[delta]}$ circuits, under the additional condition that one of the summands at the top $Sigma$ gate is squarefree. Our techniques are purely algebro-geometric: they do not rely on Sylvester--Gallai-type theorems, and our PIT result holds over arbitrary fields. The core of our proof is based on the normalization of algebraic varieties. Specifically, we carry out the analysis in the integral closure of a coordinate ring, which enjoys better algebraic properties than the original ring.