Fuzzy Cognitive Map Learning Based on Multi-Objective PSO (Invited Paper)

As a powerful paradigm for knowledge representation and causal inference, Fuzzy Cognitive Map (FCM) has gradually emerged as a powerful modeling and simulation mechanism applicable to numerous research and application fields. However, conventional FCM theory greatly depends on the experts’ knowledge. The excessive subjective factors involved in the determination of FCM weights restrict accuracy and reliability of inference results generated by FCMs. A promising approach to reducing or even eliminating the subjective intervention is the development of learning algorithm for FCMs, namely FCM learning. This paper proposes a new learning algorithm for FCMs which is based on the application of multi-objective particle swarm optimization. The novel approach integrates the FCM learning with the inference mechanism of FCMs. In order to validate the proposed FCM learning algorithm, we explore it to model the mental and physical behaviors of an emotional agent in a virtual world. The simulation results show that the novel method not only implements inference process and FCM learning in parallel, but also overcomes some deficiencies of other learning algorithms, therefore, improves the efficiency and robustness of FCMs. Copyright c © 2008 Yang’s Scientific Research Institute, LLC. All rights reserved.