Estimation of Distribution Algorithm
Estimation of Distribution Algorithms (EDA) ( Engl. , About: estimation of distribution ) are evolutionary algorithms , so procedures with the principles of evolutionary optimization problems solved. In the case of EDA, a probabilistic model is iteratively developed during the calculation , which estimates the desired optimum based on the samples taken . While all admissible solutions for the given problem are equally distributed in the model at the beginning , in the event of success, only the optimum is suggested at the end. The algorithm is a generalization of the genetic algorithm , which only implicitly estimates the distribution. The motivation for developing EDA was the fact that the selection of suitable parameters for classic evolutionary algorithms (such as mutation strength or population size) itself represents an optimization problem. John H. Holland suspected as early as 1975 that the dependencies of the variables to be optimized represent a starting point that evolutionary algorithms could exploit.
Individual evidence
- ↑ Pedro Larrañaga, José A. Lozano, Estimation of Distribution Algorithms: A New Tool for Evolutionary Computation : page 58