【摘要】 已有协同交互式文化算法中,用户之间采用广播方式实现信息共享,一方面要求信息共享同步,另一方面通信代价随用户规模的增长而增加。为此,借鉴多agent系统中的多agent联盟概念,提出一种协同交互式多agent文化算法。算法采用k-均值聚类方法,根据进化优势个体提取获得的用户偏好知识来对其进行动态分类,将具有相似偏好的用户划分到同一agent联盟。采用分层通信结构实现联盟之间和联盟内部的信息共享,特别是属于同一联盟的用户通过广播方式共享联盟共性知识。通过在购物卡配色优化设计中应用,实验结果表明所提算法能够通过有效提高进化收敛速度来较好缓解用户疲劳,并保证用户能找到满意配色方案。更多还原

【Abstract】 In existing cooperative interactive cultural algorithm,the information is shared among users via broadcast method,in which the shared information must be synchronous.This makes the communication cost increase with the increase of the number of users participating in the evolution.Therefore,by introducing the definition of multi-agent alliance in multi-agent system,cooperative multi-agent interactive cultural algorithm was presented.All users are dynamically classified by k-mean clustering method in terms of their preference.The users with the same preference belong to one alliance.The hierarchical communication structure is adopted to share common knowledge in certain multi-agent alliance and among multi-agent alliances.Especially,common knowledge is shared among users in one multi-agent alliance via broadcast method.This algorithm is applied to realize harmonious color design for shopping cards.Experimental results show that the proposed algorithm can alleviate human fatigue by improving convergence speed and ensure all users can find their satisfied painted shopping cards.更多还原