English

Universal approximation and model compression for radial neural networks

Machine Learning 2023-02-17 v3 Representation Theory

Abstract

We introduce a class of fully-connected neural networks whose activation functions, rather than being pointwise, rescale feature vectors by a function depending only on their norm. We call such networks radial neural networks, extending previous work on rotation equivariant networks that considers rescaling activations in less generality. We prove universal approximation theorems for radial neural networks, including in the more difficult cases of bounded widths and unbounded domains. Our proof techniques are novel, distinct from those in the pointwise case. Additionally, radial neural networks exhibit a rich group of orthogonal change-of-basis symmetries on the vector space of trainable parameters. Factoring out these symmetries leads to a practical lossless model compression algorithm. Optimization of the compressed model by gradient descent is equivalent to projected gradient descent for the full model.

Keywords

Cite

@article{arxiv.2107.02550,
  title  = {Universal approximation and model compression for radial neural networks},
  author = {Iordan Ganev and Twan van Laarhoven and Robin Walters},
  journal= {arXiv preprint arXiv:2107.02550},
  year   = {2023}
}

Comments

44 pages

R2 v1 2026-06-24T03:55:44.598Z