On the number of regions of piecewise linear neural networks

DOI10.1016/j.cam.2023.115667arXiv2206.08615MaRDI QIDQ6145187

Michael Unser, Arian Etemadi, Alexis Goujon

Publication date: 30 January 2024

Published in: Journal of Computational and Applied Mathematics (Search for Journal in Brave)

Full work available at URL: https://arxiv.org/abs/2206.08615

zbMATH Keywords

splinesexpressivityactivation functionsdeep learningconvex partitionscontinuous and piecewise-linear functions

Mathematics Subject Classification ID

Numerical computation using splines (65D07) Artificial neural networks and deep learning (68T07) Spline approximation (41A15)

Cites Work

• Unnamed Item
• A representation method for PWL functions oriented to parallel processing
• Region configurations for realizability of lattice piecewise-linear models.
• Spaces of convex \(n\)-partitions
• Deep vs. shallow networks: An approximation theory perspective
• Learning Deep Architectures for AI
• Generalization of Hinging Hyperplanes
• Facing up to arrangements: face-count formulas for partitions of space by hyperplanes
• A Framework for the Construction of Upper Bounds on the Number of Affine Linear Regions of ReLU Feed-Forward Neural Networks
• Sharp Bounds for the Number of Regions of Maxout Networks and Vertices of Minkowski Sums
• Stable parameterization of continuous and piecewise-linear functions
• Approximation of Lipschitz Functions Using Deep Spline Neural Networks

This page was built for publication: On the number of regions of piecewise linear neural networks