Abstract
In the system to search for a target image using IGA (Interactive Genetic Algorithm) that we have been developing, a user first selects one individual resembling a target image most and gives it fitness 1.0, and others are assigned with fitness calculated based on the relationship between the selected individual and the others. Although this system asked a user to do only simple task that is selecting one sample, all users weren't happy. There are a few reasons for this. It is sometimes difficult for a user to select only one individual. Sometimes, a couple of individuals may equally resemble a target image, or no individual may have a close resemblance to a target image. In these situations, a user wants to select a few individuals, or to say it doesn't resemble a target image very much, or it only slightly resembles a target image. In this paper, we propose a interaction strategy that is able to deal with the situations and evaluated the validity of the strategy using computer simulation. As a result, we found out useful an interaction method and a fitness assignment method.
