Extracting Building Blocks in Human-based Evolutionary Computation

Abstract

Evolutionary computation (EC), which is an optimization method inspired by biological genetics and evolution, has successfully been applied in various fields in recent years. When humans perform all evolutionary operators such as selection, crossover, and mutation for obtaining better solutions, it is called human-based EC. In a human-based EC system, humans can take the initiative in improving the performance of a solution, but it is essential to support them by computer to improve the search performance. For instance, we often combine multiple original solutions to generate a better solution, but the number of solutions is often huge. Due to that, it is often costly and hard to generate better solutions. Therefore, we propose a method for supporting participants to generate better new solutions. The method extracts building blocks (important words) in the problem solving with human-based EC and presents them. The experimental results show the basic usefulness of the method.