A Generation Gap Model for a Human-based Evolutionary Algorithm Using a Tag Cloud

We previously proposed a human-based evolutionary algorithm (human-based EA) that used a tag cloud, and in this paper, we propose a generation gap model for the human-based EA. In the previous model, people created candidate solutions, and the created candidate solutions were displayed as tags in a tag cloud. People then evaluated the candidate tags, and those with the greatest fitness were displayed using a larger font size. However, in the previous model, there was a limit on the number of tags that could be displayed in the tag cloud, and after that number had been reached, if new tags were created, they would not be included. Therefore, we propose a generation gap model for solving this problem by allowing each subsequent generation to include all of the tags from the previous generation that have a fitness level that is greater than a given threshold. The remainder in the tag cloud are replaced by tags in a queue for storage that were created before and have never displayed. We carried out an experiment using the human-based EA with the generation gap model, and the results showed that almost all participants believed the tags created by our system were better than those that they would have created by themselves.