English

Approximating Positive Homogeneous Functions with Scale Invariant Neural Networks

Machine Learning 2023-08-08 v1 Neural and Evolutionary Computing Machine Learning

Abstract

We investigate to what extent it is possible to solve linear inverse problems with ReLuReLu networks. Due to the scaling invariance arising from the linearity, an optimal reconstruction function ff for such a problem is positive homogeneous, i.e., satisfies f(λx)=λf(x)f(\lambda x) = \lambda f(x) for all non-negative λ\lambda. In a ReLuReLu network, this condition translates to considering networks without bias terms. We first consider recovery of sparse vectors from few linear measurements. We prove that ReLuReLu- networks with only one hidden layer cannot even recover 11-sparse vectors, not even approximately, and regardless of the width of the network. However, with two hidden layers, approximate recovery with arbitrary precision and arbitrary sparsity level ss is possible in a stable way. We then extend our results to a wider class of recovery problems including low-rank matrix recovery and phase retrieval. Furthermore, we also consider the approximation of general positive homogeneous functions with neural networks. Extending previous work, we establish new results explaining under which conditions such functions can be approximated with neural networks. Our results also shed some light on the seeming contradiction between previous works showing that neural networks for inverse problems typically have very large Lipschitz constants, but still perform very well also for adversarial noise. Namely, the error bounds in our expressivity results include a combination of a small constant term and a term that is linear in the noise level, indicating that robustness issues may occur only for very small noise levels.

Keywords

Cite

@article{arxiv.2308.02836,
  title  = {Approximating Positive Homogeneous Functions with Scale Invariant Neural Networks},
  author = {Stefan Bamberger and Reinhard Heckel and Felix Krahmer},
  journal= {arXiv preprint arXiv:2308.02836},
  year   = {2023}
}

Comments

31 pages

R2 v1 2026-06-28T11:48:49.281Z