CUDA-based PSO-trained neural network for computation of resonant frequency of circular microstrip antenna
by Feng Chen; Yu-bo Tian
International Journal of Computational Science and Engineering (IJCSE), Vol. 14, No. 3, 2017

Abstract: Resonant frequency is an important parameter in the design process of microstrip antenna (MSA). Artificial neural network (ANN) trained by particle swarm optimisation (PSO) algorithm (PSO-ANN) has been used to model the resonant frequency of circular MSA. In order to deal with the problem of long calculation time when training PSO-ANN, its parallel scheme in the graphic processing unit (GPU) environment is presented in this paper. The designed parallel PSO-ANN algorithm uses the particle behaviour parallelisation of PSO, corresponds one particle to one thread, and deals with a large number of GPU threads in parallel to reduce training time. This scheme is applied to model the resonant frequency of circular MSA under compute unified device architecture (CUDA). Experimental results show that compared with CPU-based sequential PSO-ANN, GPU-based parallel PSO-ANN has obtained more than 340 times speedup ratio with the same optimisation stability. Moreover, the modelling error can be remarkably reduced with a very limited runtime increment while substantially enlarging the number of particles on GPU side.

Online publication date: Tue, 16-May-2017

The full text of this article is only available to individual subscribers or to users at subscribing institutions.

 
Existing subscribers:
Go to Inderscience Online Journals to access the Full Text of this article.

Pay per view:
If you are not a subscriber and you just want to read the full contents of this article, buy online access here.

Complimentary Subscribers, Editors or Members of the Editorial Board of the International Journal of Computational Science and Engineering (IJCSE):
Login with your Inderscience username and password:

Username:        Password:          Forgotten your password?

Want to subscribe?
A subscription gives you complete access to all articles in the current issue, as well as to all articles in the previous three years (where applicable). See our Orders page to subscribe.

If you still need assistance, please email subs@inderscience.com