Dirk Thierens

Date: 16:00 – 16:30, Thursday, 20.01.2022
Location: MS Teams ICS Colloquium

Title: Model-Based Evolutionary Algorithms = machine learning + evolutionary computation

Abstract:

In classical evolutionary algorithms the search process is driven by stochastic variation operators: crossover and mutation. In model-based evolutionary algorithms (MBEAs) the variation operators are guided by the use of a model that conveys problem-specific information so as to increase the chances that exploiting the currently available solutions leads to novel, improved solutions. Such models can be constructed beforehand for a specific problem, or they can be learnt during the optimization process. Replacing traditional crossover and mutation operators by building and using models enables the use of machine learning techniques for automatic discovery of problem regularities and subsequent exploitation of these regularities, thereby enabling the design of optimization techniques that can automatically adapt to a given problem. This is especially useful when considering optimization in a black-box setting where very little a priori knowledge of the problem structure is available. We present and discuss Optimal Mixing EAs, a specific class of MBEAs.