基于人工蜂群算法的手写数字聚类研究  

The Research of Handwritten Digit Cluster Based on Artificial Bees Colony Algorithm

作　　者：王光彪[1] 杨淑莹[1] 冯帆[1] 王博凯[1] 贾紫娟[1] 朱光 

机构地区：[1]天津理工大学,计算机与通信工程学院,天津 [2]山东省广播电影电视局,山东

出　　处：《光电子》2011年第2期33-38,共6页Optoelectronics

基　　金：国家863计划项目(2007AA01Z188);国家自然基金项目(60773073);天津市高等学校科技发展基金(20071308)。

摘　　要：手写数字聚类是模式识别研究中的一个重要研究方向,但应用传统的进化算法对手写数字进行聚类分析往往存在着收敛速度慢,易陷入局部最优等问题,本文提出了用蜂群算法求解数字聚类问题,并且提出了3种蜜蜂的位置更新算子,建立了3种算子的动态更新公式,最后阐述了利用该算法对手写数字聚类的具体步骤。通过典型的手写数字实例进行了仿真实验,实验表明:该算法能够很好的实现手写数字聚类,并且克服了过早收敛的现象,而且能够加快收敛速度。Handwritten digit cluster is an important study of pattern recognition, because of the application of traditional evolutionary algorithm for clustering analysis of handwritten digit has much more problems, such as slow convergence, easily fall into local optimization and so on. To overcome those problems, we present a novel approach to solve the problem of digital clustering by using artificial bees colony algorithm, and propose 3 kinds of operators for bees’ location of updating, and establish a dynamic update of 3 operators in formulas. Finally we elaborate the concrete steps of handwriting digital cluster by using this approach. We do the simulation experiments with some typical handwritten digital instances. Experiments show that our approach can make a good implementation of handwriting digital cluster, overcome the phenomenon of pre- mature convergence, and accelerate the convergence rate in a way.

关 键 词：人工蜂群算法 手写数字聚类 组合优化 

分 类 号：TP39[自动化与计算机技术—计算机应用技术]