English

An unfeasability view of neural network learning

Machine Learning 2022-01-05 v1 Computational Complexity

Abstract

We define the notion of a continuously differentiable perfect learning algorithm for multilayer neural network architectures and show that such algorithms don't exist provided that the length of the data set exceeds the number of involved parameters and the activation functions are logistic, tanh or sin.

Keywords

Cite

@article{arxiv.2201.00945,
  title  = {An unfeasability view of neural network learning},
  author = {Joos Heintz and Hvara Ocar and Luis Miguel Pardo and Andres Rojas Paredes and Enrique Carlos Segura},
  journal= {arXiv preprint arXiv:2201.00945},
  year   = {2022}
}
R2 v1 2026-06-24T08:39:20.610Z