摘要： 介绍模拟谐振子算法，并分析其全局收敛性。将算法的进化过程分解为产生新解、修正当前解、生成新解集3个基本的进化操作，并将这种状态变化分别映射为3个随机矩阵。应用有限马尔科夫链理论对该算法的解状态矩阵变化进行分析，结果表明，在保留优质解的前提下，当运算时间趋于无穷时，算法会逐渐收敛于全局最优解。

关键词: 智能计算, 模拟谐振子, 有限马尔科夫链, 随机矩阵, 状态转移概率, 全局收敛

Abstract: This paper introduces the Simulated Harmonic Oscillator(SHO) algorithm, and analyses the global convergence of it. Process of SHO is divided into three basic operations, such as generating new solution, amending current solution, composing new solution sets, and mapping these changes of state into three stochastic matrixs. Using limited Markov chain theoretics to analyse the matrix of state changing, it is proved that when the running time goes infinity, SHO with keeping excellent answer can keep the best global solution convergent.

Key words: intelligent computing, Simulated Harmonic Oscillator(SHO), limited Markov chain, stochastic matrix, state transition probability, global convergence

中图分类号: 

• TP301.6