双层进化交互式遗传算法的知识提取与利用

Extraction and utilization about knowledge in hierarchical interactive genetic algorithms

针对交互式遗传算法缺乏知识利用的通用结构框架问题,借鉴文化算法的双重进化结构,提出一种交互式遗传算法中知识提取及利用的通用结构框架.构建了一种反映人认知和偏好等隐含知识,融合常识、进化知识和评价知识的广义知识模型.利用迁移分析方法证明了算法的收敛性,给出了近似模型替代人评价的临界代数.以服装进化设计系统为例,验证了算法结构和知识模型的合理性.仿真实例与分析结果表明,所提出的算法可以有效缓解人的疲劳,提高进化收敛速度.

For the problem that interactive genetic algorithms lack a universal frame to utilize knowledge, a universal frame for extraction and utilization for knowledge in interactive genetic algorithms is proposed by adopting dual structure in culture algorithms. A knowledge model composed of common sense, evolution knowledge and evaluated knowledge is constructed, which describes implicit knowledge about users＇ cognitive and preference. Convergence is proved by using drift analysis, and critical generation substituting approximate model for users＇ evaluation are achieved. Based on fashion evolutionary design system, the rationality of this algorithm and the validity of the knowledge model are proved. Simulation results indicate that the algorithm can effectively alleviate users＇ fatigue and improve the speed of convergence.