← 研踪 PaperTrailarXiv:2509.05894 · v2
Toric geometry of ReLU neural networks
首次提交 2025-09-07 10:08
版本更新 2026-11-13 09:32
雷达观测 2026-11-17 04:57
摘要
Given a continuous piecewise linear function and a fixed architecture of feedforward ReLU neural networks, the exact function realization problem asks when a network of that architecture realizes . In contrast, expressivity concerns characterizing all functions exactly realizable by a given architecture. To address these questions systematically, we establish a connection between toric geometry and ReLU neural networks, allowing tools from algebraic geometry to analyze the structures. Starting with an unbiased ReLU network with rational weights, we define the ReLU fan, the ReLU toric variety, and the ReLU Cartier divisor associated with the network. This work also relates the tropical and the toric geometry: the former captures combinatorial data of ReLU networks, while the latter recovers the underlying geometric structure. As an application, we obtain a complete characterization of the functions realizable by unbiased shallow ReLU networks with rational weights. The characterization is expressed in terms of intersection numbers of ReLU Cartier divisors with torus-invariant curves and leads to a decomposition of piecewise-linear functions into shallow and non-shallow components.
观测到的版本
相邻记录可能跳过中间版本;首次发现 v2 及以上也会标为版本更新。
v2 · 2026-11-13 09:32 · 最新观测版本
Toric geometry of ReLU neural networks
Given a continuous piecewise linear function and a fixed architecture of feedforward ReLU neural networks, the exact function realization problem asks when a network of that architecture realizes . In contrast, expressivity concerns characterizing all functions exactly realizable by a given architecture. To address these questions systematically, we establish a connection between toric geometry and ReLU neural networks, allowing tools from algebraic geometry to analyze the structures. Starting with an unbiased ReLU network with rational weights, we define the ReLU fan, the ReLU toric variety, and the ReLU Cartier divisor associated with the network. This work also relates the tropical and the toric geometry: the former captures combinatorial data of ReLU networks, while the latter recovers the underlying geometric structure. As an application, we obtain a complete characterization of the functions realizable by unbiased shallow ReLU networks with rational weights. The characterization is expressed in terms of intersection numbers of ReLU Cartier divisors with torus-invariant curves and leads to a decomposition of piecewise-linear functions into shallow and non-shallow components.
查看此版本解读 →查看此版本原文 ↗