English

A view of Estimation of Distribution Algorithms through the lens of Expectation-Maximization

Machine Learning 2022-06-14 v9 Machine Learning

Abstract

We show that a large class of Estimation of Distribution Algorithms, including, but not limited to, Covariance Matrix Adaption, can be written as a Monte Carlo Expectation-Maximization algorithm, and as exact EM in the limit of infinite samples. Because EM sits on a rigorous statistical foundation and has been thoroughly analyzed, this connection provides a new coherent framework with which to reason about EDAs.

Keywords

Cite

@article{arxiv.1905.10474,
  title  = {A view of Estimation of Distribution Algorithms through the lens of Expectation-Maximization},
  author = {David H. Brookes and Akosua Busia and Clara Fannjiang and Kevin Murphy and Jennifer Listgarten},
  journal= {arXiv preprint arXiv:1905.10474},
  year   = {2022}
}