Abstract
An improved differential evolution (DE) algorithm based on global sensitivity analysis is proposed to enhance performance in high dimension problems with limited computation resources. Morris-One-at-a-Time(MOAT) method was firstly tested on a typical function and compared with Sobol sensitivity method, showing high efficiency and acceptable result. Then MOAT method is used to calculate sensitivity for each dimension of input vector, and new crossover and mutation operators are proposed to incorporate sensitivity for two improved algorithm GSADE1 and GSADE2. Five 50-dimension functions were used for test, showing both two new algorithms are better than the original DE.
| Original language | English |
|---|---|
| Pages (from-to) | 411-417 |
| Number of pages | 7 |
| Journal | Xibei Gongye Daxue Xuebao/Journal of Northwestern Polytechnical University |
| Volume | 34 |
| Issue number | 3 |
| State | Published - 1 Jun 2016 |
Keywords
- Algorithm
- Computer simulation
- Differential evolution
- Global optimization
- Sensitivity analysis