Estimation of Distribution Algorithms: A Recent Approach to Evolutionary Computation

Evolutionary algorithms (EAs) have become one of the successful optimization methods during the last few decades, especially for problems where smooth optimization cannot be used. This work summarises present trends in Estimation of distribution algorithms (EDAs), a recent kind of EAs, and suggests a further research direction. These algorithms evolve a population of partial solutions, and instead of modifying the individuals by genetic operations, they construct the probability distribution of their variables and use it for sampling new population. Versions of EDAs for discrete and continuous parameters are presented as well as different approaches to construction of probability distributions including the very recent usage of copulas.