English

Learning Narrow One-Hidden-Layer ReLU Networks

Machine Learning 2023-04-21 v1 Data Structures and Algorithms Machine Learning

Abstract

We consider the well-studied problem of learning a linear combination of kk ReLU activations with respect to a Gaussian distribution on inputs in dd dimensions. We give the first polynomial-time algorithm that succeeds whenever kk is a constant. All prior polynomial-time learners require additional assumptions on the network, such as positive combining coefficients or the matrix of hidden weight vectors being well-conditioned. Our approach is based on analyzing random contractions of higher-order moment tensors. We use a multi-scale analysis to argue that sufficiently close neurons can be collapsed together, sidestepping the conditioning issues present in prior work. This allows us to design an iterative procedure to discover individual neurons.

Keywords

Cite

@article{arxiv.2304.10524,
  title  = {Learning Narrow One-Hidden-Layer ReLU Networks},
  author = {Sitan Chen and Zehao Dou and Surbhi Goel and Adam R Klivans and Raghu Meka},
  journal= {arXiv preprint arXiv:2304.10524},
  year   = {2023}
}

Comments

33 pages, comments welcome

R2 v1 2026-06-28T10:12:52.799Z