Estimation of distribution algorithm
From Wikipedia, the free encyclopedia
In evolutionary computation the population may be approximated with a probability distribution over the space of possible solutions. This may have several advantages, including avoiding premature convergence and being a more compact representation.
Better known EDAs include the Compact Genetic Algorithm, the Population Based Incremental Learning and the Univariate Marginal Distribution Algorithm.
The model may be found to fit an existing population or take on the role of the population entirely. Once the model is obtained, it can be sampled to produce more candidate solutions which are then used to adapt or regenerate the model. EDAs are typically classified according to the level of variable interaction that their probabilistic model includes - they can be classed as univariate (no interactions), bivariate (interactions between pairs of variables) or multivariate (interactions between more than two variables).
