论文部分内容阅读
针对二阶分布估计算法的早熟收敛问题,提出一种基于混合采样机制的互信息分布估计算法(MIEDA).MIEDA利用互信息度量变量之间的相关性,形成互信息树的概率模型;采用稀疏模型构建的思想,并基于自私基因理论建立信息奖惩机制,以加快算法的收敛速度;结合反向学习、最优解变异和随机采样形成混合采样机制,以提高算法的采样效率.仿真结果表明,MIEDA比常见的二阶分布估计算法具有更高的稳定性和更强的寻优能力.
Aiming at the premature convergence problem of second-order distribution estimation algorithm, a hybrid sampling algorithm based mutual information distribution estimation algorithm (MIEDA) is proposed.MIEDA uses the mutual information to measure the correlation between variables to form the mutual information tree probability model, Model construction and set up information rewards and punishments mechanism based on selfish gene theory to speed up the convergence of the algorithm.A combination of reverse learning, optimal solution mutation and random sampling is used to form a mixed sampling mechanism to improve the sampling efficiency of the algorithm.The simulation results show that, MIEDA than the common second-order distribution estimation algorithm has a higher stability and better search ability.