摘要
基于Bayes决策理论,提出了一种可以改进蚁群算法搜索性能的有效方法;针对基本蚁群算法中存在的“停滞”现象,对蚂蚁个体的寻优过程采取了隔代强化的措施,使算法具备较强的发现新解的能力,再采用后验分析对蚁群算法中的转移概率进行调整,使得改进后的蚁群算法在随机搜索过程中呈现出自组织特性,蚂蚁个体利用各自的后验知识不断地强化那些能“经受考验”的可行解,从而有效地压缩了搜索空间,提高了搜索效率.试验结果表明,该方法无需知道转移概率的先验分布,在解空间的全局寻优时具有良好的收敛性和鲁棒性.
An effective method based on the principle of Bayes decision is put forth to improve the searching performance of basic ant colony algorithms. Aiming at the stagnation phenomenon, a way of interval strengthening is applied, thus the new method has a good ability of finding new solution.Meanwhile, posterior analysis is adopted to adjust the diversion probability and is applied by each agent to strengthen those durable solutions, which makes the stochastic searching process of the modified algorithms appear self-organizing characteristics and reduce the hunting sphere largely and improve the searching efficiency. The results of experiment show that the proposed method, even without any knowledge of diversion probability＇s prior distribution, has favorable convergence and robustness in finding the optimal solution.
出处
《东南大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2005年第4期558-562,共5页
Journal of Southeast University：Natural Science Edition
基金
国家自然科学基金资助项目(69875004)
江苏省自然科学基金资助项目(BK2001402).
关键词
蚁群算法
Bayes决策
极大熵
先验分布
后验分析
ant colony algorithms
Bayes decision
maximum entropy
prior distribution
posterior analysis
