A Coevolution Approach for Learning Multimodal Concepts

In this paper, we propose a cooperative coevolution approach to learn rules for the description of multimodal concepts. Multiple species are evolved in parallel; each evolves a particular part of the description of the target concepts. The algorithm allows the number and roles of the species to be adapted; more accurate and general rules are generated. The proposed algorithm has been compared to other two popular concept learning algorithms on five benchmark datasets from the UCI machine learning repository. Results show that the proposed algorithm can achieve higher performance while still produces a smaller number of rules.