Optimization via Classification

The vast majority of population-based optimization algorithms use selection in such a way that the nonselected individuals do not have any effect on the evolution at all, even though they may carry a valueable information — information about the local shape of the search distribution and/or about the search space areas where the search should be suppressed. This article describes a unified way of taking advantage of the information hidden in the non-selected individuals in the framework of evolutionary algorithms: first, build a classifier discriminating between selected and non-selected individuals, then turn the description of selected individuals into a search distribution, and sample new offspring from it. The concept is verified by a simple real-valued evolutionary algorithm which outperforms the state-of-the-art evolutionary strategy with covariance matrix adaptation (CMA-ES) on selected test functions in all tested search space dimensionalities. Finally, the article proposes some guidelines for future work to make this algorithm generally applicable.