Estimation of Distribution Algorithm

Estimation-of-distribution algorithms (EDAs) are optimization methods that learn probabilistic models of promising solutions, iteratively refining these models to efficiently explore the search space. Current research emphasizes extending EDAs beyond binary problems to handle multi-valued variables and analyzing the impact of key parameters like population size and model update strength, particularly concerning the phenomenon of genetic drift. This work aims to improve the efficiency and robustness of EDAs, leading to better performance on complex optimization problems across various domains. A unified theoretical framework for understanding different EDA variants is also emerging, facilitating more general analyses and potentially informing the design of more efficient algorithms.