Designing Particle Swarm Optimization - Performance Comparison of Two Temporally Cumulative Fitness Functions in EPSO

H. Zhang and M. Ishikawa (Japan)


particle swarm optimization, real-coded genetic algorithm, elitism strategy, temporally cumulative fitness function, model selection


We present an Evolutionary Particle Swarm Optimization (EPSO) method for PSO model selection. It provides a new paradigm of meta-optimization that systematically es timates appropriate values of parameters in PSO for ef ficiently finding an optimal solution to a given optimiza tion problem. For investigating the characteristics, i.e., ex ploitation and exploration of the optimized PSO, this paper proposes to use two fitness functions in EPSO, which are a temporally cumulative fitness of the best particle and a temporally cumulative fitness of the entire swarm. Appli cations of the proposed method to a 2-dimensional opti mization problem well demonstrate its effectiveness. The obtained results indicate that the former fitness function can generate a PSO model with higher fitness, and the latter can generate a PSO model with faster convergence.

Important Links:

Go Back