Developing Social Informational Network Models Based on Evolutionary Computation and Machine Learning Theories

该研究项目的目标是开发基于代理的模型，以揭示现代社会信息网络，如互联网的复杂和新兴的属性。在基于多智能体的复杂适应系统方法框架中，我们利用了进化计算和机器学习理论。（1）面向组织学习的分类器系统（OCS）OCS是遗传算法中传统的学习分类器系统在多智能体和组织学习环境中的一种新的扩展。(2)逆向仿真技术及其在社会Agent仿真器TRURL中的应用逆向仿真是一种分析复杂仿真任务中现象的新技术。这是实现新的遗传算法的多峰和多目标函数优化。(3)应用的技术，以实际的任务域所开发的方法被应用到实际的任务域：印刷电路板零件放置，Fairy-Wing：智能IC卡引导系统，协同进化在生态营销模拟，知识管理和社会互动研究。

The objective of the research project is to develop agent-based models to uncover the sophisticated and emergent properties of modern social information networks such as the Internet. We have utilized both evolutionary computation and machine learning theories in the framework of multiagent-based complex adaptive system approach. The main results are summarized as follows :(1) Organizational Learning Oriented Classifier System (OCS)OCS is novel extension of conventional Learning Classifier Systems in the GA literature to multiagent and organizational learning environments.(2) Inverse Simulation Techniques and its Application to Social Agent-based Simulator TRURLInverse Simulation is a novel technique to analyze phenomena in complex simulation tasks. This is attained by new GAs for multimodal and multiobjective function optimization.(3) Application of the Techniques to Practical Task DomainsThe developed methods are applied to practical task domains : Printed Circuit Board Parts Placement, Fairy-Wing : Guidance System with Smart IC-Cards, Coevolution in Ecomarketing Simulation, Knowledge Management and Social Interaction Studies.