基于广义量子粒子模型的聚类算法及收敛性研究  

Research on a Clustering Algorithm Based on Generalized Quantum Particle Model and its Convergence

作　　者：黄良军[1] 帅典勋[1,2] 张彬[1] 

机构地区：[1]华东理工大学计算机科学与技术系,上海200237 [2]清华大学智能技术与系统国家重点实验室,北京100084

出　　处：《计算机科学》2010年第2期225-228,236,共5页Computer Science

基　　金：国家自然科学基金(60575040;60473044)资助

摘　　要：提出了一种应用广义量子粒子模型进行自组织聚类的新方法。该模型将数据聚类过程转化为一个量子粒子在状态构形空间上的随机自组织过程,由量子粒子之间相互纠缠形成的状态构形随时间不断演化,最终会收敛到一个平稳的概率分布,最优状态空间构形与平稳概率分布中具有最大概率的状态构形相对应。对此自组织过程的收敛性进行了理论上的证明。与传统的适用于大规模数据的聚类方法相比较,该算法具有更快的收敛速度,仿真实验表明了其优越性。A novel generalized quantum particle model （GQPM） was presented for data self-organizing clustering. In this model the data clustering process is transformed into a stochastic self-organizing process of the quantum particles in the state configuration space. The state configuration will evolve to a stationary probability distribution, and thus the optimal state configuration on particles can be obtained from the state configuration which has the highest probability in the stationary probability distribution. The convergence of the self-organizing process was proved in this paper. The GQPM algorithm has much faster clustering speed than the traditional clustering algorithm for the large scale database. Its superiorities were verified by the simulation experiments.

关 键 词：数据聚类 高维数据 随机过程 马尔可夫链 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]