混沌神经网络组合优化计算及其应用-系统工程专业论文.docxVIP

摘要下降退火网络混沌神经网络组合算法组合优化问题TSP 摘要 下降退火网络混沌神经网络组合算法组合优化问题TSP AbstractABSTRACT Abstract ABSTRACT Several chaotic neural networks for solving combinatorial optimization problems are studied in this dissertation．The main research is as follows．A combined algorithm is proposed，using the the transiently chaotic neural network(TCⅫ叼and a local search algorithm，which is able to find the optimal solution of Chinese traveling salesman problem(CTSP)．Nonchaotic simulated annealing of TCNN is proposed， which is different from chaotic simulated annealing proposed by Chen and Aihara．An improved version of TCNN is proposed by using linear activation function and the improvement is great．An algorithm which is good as simulated annealing and ant algorithm is proposed by combining the chaotic neural network and the 2-opt algorithm．The solving ability of this algorithm for combinatorial optimization problems is proved by solving several TSPs． In charpter one and two，the histories and the models of the Hopfield neural networks，the chaotic neural networks and TCNN are expatiated．And the principle and mechanism of these three models are analyzed． In charpter three，a combined algorithm which uses the transiently chaotic neural network and a local search algorithm is proposed．1he combined algorithm searches further based on the results acquired by TCNN．Tests on CTSP show that the new method can get the optimal solution of CTSP．And also，nonchaotic simulated annealing is introduced in this charpter．Nonchaotic simulated annealing is obtained by altering the initial value of self-feedback weight of TCNN．Tests and comparisons on the traveling salesman problem(TSP)show the improvements of nonchaotic simulated annealing over chaotic simulated annealing．A lot of computing time needed of TCNN iS saved． In charpter four,an improved version of the transiently chaotic neural network is proposed．The linear activation function is used instead of the original sigmoid activation function．The new model is calle