Study On Some Stochastic Global Optimization Algorithms

Many problems in practically all fields of natural science as well as technical design, financial planning and travel scheduling involve global optimization problems. The efficient global optimization methods affect the development of these subjects. With the development of the stochastic global optimization method such as tabu search, simulated annealing and evolutionary computation during 80 age in the 20th century, some authors study the theory and application of those algorithms and put forward novel algorithms and solve a lot of practical problems. Based on the former research, we study the tabu search, simulated annealing and evolutionary programming in the aspects of theory, application and improvement. The paper is organized as follows:In Chapter 1, we introduce in detail the status of tabu search, simulated annealing and evolutionary computation in the aspects of theory, improvement and application. In addition, we introduce the main research of my paper.In Chapter 2, we put forward the tabu search based on memory. In theory, the algorithm converges to the global minimum point with probability one under suitable condition. Numeral results show that tabu search based on memory is superior to the classical stochastic methods such as tabu search, simulated annealing and so on. Otherwise, we apply tabu search based on memory to solving the continuous global optimization problem of medical image registration and had good experimental results.In Chapter 3, we propose a kind of simulated annealing, which contains memory simulated annealing, chaos simulated annealing and improved simulated annealing. We prove memory simulated annealing converge to the global minimum point with probability one under suitable condition. We apply chaotic systems to simulated annealing and propose chaos simulated annealing. Numeral results illustrate that chaos simulated annealing is efficient and easy to implement. we propose a improved simulated annealing which can solve the linear constrained optimization problems and good results have been given.In Chapter 4, we propose a novel evolutionary programming, which named single-point mutation evolutionary programming. Numeral results shows that the algorithm is superior to the classical evolutionary programming, fast evolutionary programming and generalized evolutionary programming for high-dimensional and multi-modal functions. The algorithm spends much less CPU time and have higher robust than others. In addition, we generalize the algorithm to solve linear constrained optimization problems and give good results.In Chapter 5, we conclude my research and present my further research in the future.