| Summary: | 碩士 === 國立金門技術學院 === 電資研究所 === 98 === Hill-Climbing(HC) is an optimization algorithm that is widely known. It’s simple and fast. However, HC will be trapped when it fall into local optima, and cannot find global optimal solution. In this paper, we design a jumping strategy to help HC escape from local optima. The new algorithm is called Hill-Climbing with Jumping Strategies(HCJ).
This text adopts three popular Nonlinear Programming problems - G1, G7, G9, to test the effect of HCJ algorithm. The result showed that HCJ had excellent performances on G1, G9, but HCJ couldn't work well on G7. Through analyzing the experiment, we realized that G7 may contain a problem itself, this is a problem which HCJ cannot solve at the moment, but it is also an area which HCJ could be improved in the future.
However, when comparing HCJ with other single-particle algorithm, such as HC and SA, HC shows much improvement in its accuracy with its performance and solutions; when compared with multi-particle algorithm, we have found out that when HCJ is obtaining a solution, although it is not as good as the modified PSO and GA, but it is better than the original PSO and GA. This is resulted from the modification of the HCJ jumping strategy, it also shows the improvement in the potential of the HCJ jumping strategy.
|
|---|
