English

Approximation Capabilities of Neural ODEs and Invertible Residual Networks

Machine Learning 2020-03-03 v2 Machine Learning

Abstract

Neural ODEs and i-ResNet are recently proposed methods for enforcing invertibility of residual neural models. Having a generic technique for constructing invertible models can open new avenues for advances in learning systems, but so far the question of whether Neural ODEs and i-ResNets can model any continuous invertible function remained unresolved. Here, we show that both of these models are limited in their approximation capabilities. We then prove that any homeomorphism on a pp-dimensional Euclidean space can be approximated by a Neural ODE operating on a 2p2p-dimensional Euclidean space, and a similar result for i-ResNets. We conclude by showing that capping a Neural ODE or an i-ResNet with a single linear layer is sufficient to turn the model into a universal approximator for non-invertible continuous functions.

Keywords

Cite

@article{arxiv.1907.12998,
  title  = {Approximation Capabilities of Neural ODEs and Invertible Residual Networks},
  author = {Han Zhang and Xi Gao and Jacob Unterman and Tom Arodz},
  journal= {arXiv preprint arXiv:1907.12998},
  year   = {2020}
}