The BUMDA
The Boltzmann Univariate Marginal Distribution Algorithm (BUMDA), is a direct optimization method from the family of Estimation of Distribution Algorithms (EDAs), a kind of black-box optimization methods which intend to approximate the optimum of a function by estimating and sampling from a probability distribution. The BUMDA approximates the optimum of a function defined in a n-dimensional real domain, by using a set of univariate Gaussian distributions. The parameters of each Gaussian distribution are computed to minimize the Kullback-Leibler divergence from a Boltzmann distribution which energy function is the objective function. This page provides of a draft version of the article, and matlab source code of the method. For any request or question please send it to ivvan@cimat.mx
The BUMDA full reference
- S. Ivvan Valdez, Arturo Hernández, and Salvador Botello. A Boltzmann based estimation of distribution algorithm. Information Sciences, Volume 236, 1 July 2013, Pages 126--137. doi:10.1016/j.ins.2013.02.040
Documents
The BUMDA at ScienceDirect of Elsevier (Information Sciences journal))
Source code