English

Universal Approximation of Markov Kernels by Shallow Stochastic Feedforward Networks

Machine Learning 2015-03-26 v1 Machine Learning

Abstract

We establish upper bounds for the minimal number of hidden units for which a binary stochastic feedforward network with sigmoid activation probabilities and a single hidden layer is a universal approximator of Markov kernels. We show that each possible probabilistic assignment of the states of nn output units, given the states of k1k\geq1 input units, can be approximated arbitrarily well by a network with 2k1(2n11)2^{k-1}(2^{n-1}-1) hidden units.

Cite

@article{arxiv.1503.07211,
  title  = {Universal Approximation of Markov Kernels by Shallow Stochastic Feedforward Networks},
  author = {Guido Montufar},
  journal= {arXiv preprint arXiv:1503.07211},
  year   = {2015}
}

Comments

13 pages, 3 figures

R2 v1 2026-06-22T09:01:17.132Z