对EDA算法模型地改进和探讨.docxVIP

上海交通大学硕士学位论文 对问题变量之间相互关系的假设同样可将新算法模型分为两类结 合具体的线性组和优化问题通过实验讨论了新的 EDA 算法模型在性能运算规模和收敛速度等方面的优缺点 关键词进化算法遗传算法EDA 算法选择算子变异算子 基因池重组算子 7 上海交通大学硕士学位论文 The Modification and Discussion of The Estimation of Distribution Algorithm Abstract: Evolutionary Algorithm (EA) is to establish a math model by simulating the biology evolutionary process, which is a whole optimization search algorithm. EA starts from explaining the biology’s complicated adapting process in nature system, simulates the features of biology evolution to establish the man-made system model, introducing a universal structure to solve the complex system optimization problems. Genetic Algorithm (GA) is one of EA, mainly using the genetic operation to recombine individuals with specified structure in colony, which finding the structure comparability in individuals constantly, and approaching the final optimization result gradually. Schemata theorem is the base of GA, which guarantees the sample number of better schemas (better results of GA) will increasing by index level, meets the necessary condition to find the optimization result, that is the possibility to find the whole optimization result by GA. The building block hypothesis indicates that GA has the ability to find the optimization result, that is in GA the building blocks will the high ranks, long distance, and high average adapting schema to reach the final optimization result. But the crossover operators of GA are independent to the optimization problem, which can 8 上海交通大学硕士学位论文 not be the building blocks sometimes, even destroy the building blocks frequently. Further the genetic parameters of GA are not decided easily. So now there are a lot of methods to research the structure problem, using the structure information to establish and optimize the building blocks effectively, and EDA is one of them. In EDA method, a probability model is established by the potential results collections, using it to further detect search spare of the problem. EDA Algorithm