基于模拟退火的DEAS算法

Global optimization method:SA-Based Dynamic Encoding algorithm for Searching

针对动态编码搜索算法(DEAS)求解全局优化问题容易陷入局部最优解的问题,提出一种基于模拟退火思想的动态编码随机搜索算法。算法的静态数据结构是二进制矩阵,矩阵每一行代表问题的一个维度;动态过程包括增加串长执行搜索和在最优方向的引导下探索两个基本过程。数值实验的结果表明,对非线性的和不连续的多维函数,改进随机算法的性能要优于原始DEAS算法,具有对初始解强的鲁棒性和更强的跳出局部最优解的优点。

A new stochastic Dynamic Encoding Algorithm for Searching （DEAS） inspired by the Simulate Annealing （SA） algorithm is proposed to deal the problem that the original method is inclined to be trapped in the local optimal solutions.The structure of this approach is a binary matrix in which each row represents a parameter of a corresponding problem,and the two basic processes involve with bisectional searching by increasing the length of binary strings and unidirectional searching guided by the optimal direction which is constructed by the least significant bit of the optimal matrix.The numerical simulation results show that the Sa-based approach is fairly robust to initial conditions and its performance is superior to that of single method.