13 March 2013 A novel fitness evaluation method for evolutionary algorithms
Author Affiliations +
Abstract
Fitness evaluation is a crucial task in evolutionary algorithms because it can affect the convergence speed and also the quality of the final solution. But these algorithms may require huge computation power for solving nonlinear programming problems. This paper proposes a novel fitness evaluation approach which employs similarity-base learning embedded in a classical differential evolution (SDE) to evaluate all new individuals. Each individual consists of three elements: parameter vector (v), a fitness value (f), and a reliability value(r). The f is calculated using NFEA, and only when the r is below a threshold is the f calculated using true fitness function. Moreover, applying error compensation system to the proposed algorithm further enhances the performance of the algorithm to make r much closer to true fitness value for each new child. Simulation results over a comprehensive set of benchmark functions show that the convergence rate of the proposed algorithm is much faster than much that of the compared algorithms.
© (2013) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Ji-feng Wang, Ji-feng Wang, Ke-zong Tang, Ke-zong Tang, "A novel fitness evaluation method for evolutionary algorithms", Proc. SPIE 8784, Fifth International Conference on Machine Vision (ICMV 2012): Algorithms, Pattern Recognition, and Basic Technologies, 87840E (13 March 2013); doi: 10.1117/12.2013684; https://doi.org/10.1117/12.2013684
PROCEEDINGS
7 PAGES


SHARE
Back to Top