Knowledge Learning in Interactive Evolutionary Computation Based on Information Flow

Reducing users´ fatigue and improving the performance are two focuses of the research of interactive evolutionary computation (IEC). Aiming at the focuses, knowledge learning in IEC is put forward. Before the discussion of knowledge learning, the issue of information sampled from history evolution is discussed, from which the knowledge is extracted. The knowledge learning based on gene-sense-unit (GSU) is put forward and the knowledge are mainly embodied in the function of predicting fitness, in the methods to extract user-preference. The experiments validate the efficiency of the proposed methods which can be effectively reduce user fatigue and improve the performance of the algorithm. The above research establishes necessary foundation for future study.