论文部分内容阅读
模拟退火 (SA )算法是一种常用的概率性全局优化算法 ,但其搜索行为和优化性能对参数有严重的依赖性 ,其中状态发生器的设计最为关键。论文主要研究函数优化中基于 Cauchy分布的状态发生器 (SGC)和基于 Gaussian分布的状态发生器 (SGG)对 SA算法性能的影响。对分布机制的研究表明 ,SGC有利于大范围搜索和脱离极小区域 ,而SGG较适合于局部搜索。对不同复杂度的典型问题的仿真表明 ,优化简单单极小问题时 SGC的优化效率优于基于SGG,优化复杂多极小或存在平坦区的简单问题时 SGC的优化度和鲁棒性均优于 SGG。进而利用对尺度参数的“退温”控制 ,提出了 SGC的改进策略 ,较大程度上提高了优化度和鲁棒性。
The simulated annealing (SA) algorithm is a commonly used probabilistic global optimization algorithm, but its search behavior and optimization performance have serious dependence on the parameters. The design of the state generator is the most important. The dissertation mainly studies the influence of the state generator (SGC) based on Cauchy distribution and the state generator (SGG) based on Gaussian distribution in the function optimization on the performance of SA algorithm. Research on the distribution mechanism shows that SGC is good for large-scale search and separation from very small area, while SGG is more suitable for local search. The simulation of typical problems with different complexities shows that the optimization efficiency and the robustness of SGC are superior to those of SGC based on SGG in the optimization of simple unipolar problem and the simple problem of optimization of complex multipolar or flat region At SGG. Furthermore, by using the “warm-down” control of scale parameters, an improved SGC strategy is proposed, which improves the optimization degree and robustness to a great extent.