Adaptation of optimization algorithms to problem domains by transfer learning

Optimization is one of the most important problems in science and engineering. Common optimization algorithms are designed to work for a large set of problems, but not necessarily to be efficient for specific domains. We propose a new transfer learning approach to adapt optimization algorithms to specific problem domains. Our approach analyzes solved problems of a domain to identify areas in the search space where good solutions are expected for this domain. Knowledge of these areas is used to improve the optimization algorithm performance of unseen problems of the same domain. Because of its general design, our method can be applied to a wide range of problems and algorithms.