Classification of Sonar Targets using Particle Swarm Optimization via Independent Groups

Authors

1 Department of Electrical Engineering, Iran University of Science and Technology, Tehran, Iran

2 Department of Electrical Engineering, Shiraz University of Technology, Shiraz, Iran

3 Department of Electrical Engineering, Imam Khomeini University, Noshahr, Iran

Abstract

Due to the fact that sonar targets have high dimensions and local optimums, conventional classifiers do not have adequate ability to classify these targets. Using a combination of Particle Swarm Optimization (PSO) and Artificial Neural Networks (ANNs) is one of the solutions to overcome this problem. PSO has two drawbacks in high-dimensional datasets: being trapped in local minimums and slow convergence rate. To tackle these deficiencies, this paper uses a newly proposed meta-heuristic algorithm entitled Independent Group Particle Swarm Optimization (IGPSO). This algorithm is inspired by the diversity of individuals in the accumulation of birds or the swarm of insects. It has the unique ability to classify high-dimensional dataset (sonar). In order to test the capabilities of the IGPSO, the algorithm will be evaluated by 23 well-known test functions and the results are compared to PSO and improved versions of PSO. The results show that IGPSO is able to provide much better results in finding the global minimum of functions, convergence speed and local minima avoidance compared with other benchmark algorithms. The results show that the classifier which is designed with IGPSO classifies sonar dataset with accuracy about 96.67% while the accuracy of the PSO-classifier is about 92.33%.

Keywords

Files

Share

How to cite

Statistics