Modeling Resonant Frequency of Rectangular Microstrip Antenna Using CUDA-Based Artificial Neural Network Trained by Particle Swarm Optimization Algorithm

Resonant frequency is a vital parameter in designing Microstrip Antenna (MSA). Artificial Neural Network (ANN) based on Particle Swarm Optimization (PSO) algorithm (PSO-ANN) has been used to model the resonant frequency of rectangular MSA. To deal with the problem of the long execution time when training PSO-ANN, its parallel implementation in the Graphic Processing Unit (GPU) environment is proposed in this paper. The presented approach uses the particle behavior parallelization of PSO to accelerate ANN training, and is applied to modeling the resonant frequency of rectangular MSA under Compute Unified Device Architecture (CUDA). Experimental results indicate that compared with CPU-based sequential PSO-ANN, more than 300 times of speedup ratio has achieved in GPU-based parallel PSO-ANN with the same optimization stability. Furthermore, the network error can be significantly reduced with the very limited runtime increment when substantially enlarging the number of particles on GPU side.