Research of Ranking Method in Evolution Strategy for Solving Nonlinear System of Equations
Huantong Geng, Yijie Sun, Qing-Xi Song, Tingting Wu · 2009
As new methods, Evolutionary algorithms (EA) have been adopted to solve these complicated nonlinear systems of equations (NSE) problems. Especially, during solving these NSE problems with evolution strategy (ES), the traditional evaluation method between two different individuals has some drawbacks because of not considering the conflict of the different equation objectives, and then it more likely sticks into local optimum on these problems with two or more conflicting equation objectives. Therefore, this paper proposes a new probability ranking method based on the individual fitness value when the comparing two individuals appear the different equation-difference conflict. The experimental results show that our modified algorithm holds the better performance of global searching and less probability of sticking into local optimum.