Recent Topics of Interactive Evolutionary Computation
Hideyuki Takagi (Japan)

Interactive Evolutionary Computation (IEC) is a method for optimizing target systems based on human knowledge, experiences, preference, intuition, and/or KANSE in general. There are many systems that it is hard or impossible to design fitness functions for the optimization and therefore we cannot apply conventional optimization methods including conventional evolutionary computation (EC) framework.

First, we overview several IEC applications to understand what is IEC feasibly as introduction.

Secondly, we explain recent three IEC research topics: (1) IEC for human science, (2) IEC with physiology, and (3) IEC with evolutionary multi-objective optimization (EMO). As IEC optimizes target systems based on human psychology or physiology, we may obtain unknown knowledge on human science by analyzing the optimized target system or human-EC interaction process. The second topic is extended IEC that optimizes target systems based on human physiological feedback. The third topic is handling EMO when some of objectives must be evaluated subjectively. In this case, conventional EMO approach cannot handle the tasks and needs to combine with human evaluation.

Thirdly, we explain some IEC frameworks that include (1) interactive PSO, parallel IEC, (3) tournament IEC and simulated breeding, and (4) interactive Differential Evolution. Their aim is to reduce IEC user fatigue by accelerating EC convergence or improving IEC interface and make the IEC a practical technique.