Improved Opposition-Based PSO for Feedforward Neural Network Training

Abstract

In this study we present an improved opposition- based PSO and apply it to feedforward neural network training. The improved opposition-based PSO utilizes opposition-based initialization, opposition-based generation jumping and opposition-based velocity calculation. The opposition-based PSO is first tested on some unimodal and multimodal problems and its performance is compared with standard PSO. We then test the performance of the improved opposition-based PSO for training feedforward neural network and also present a comparison with standard PSO.