Contribution à la modélisation et à l’optimisation de structures et dispositifs microondes en utilisant divers types de réseaux de neurones.

Abstract

In this work, a new strategy of neural networks (NN) is proposed to modeling microwave waveguide structures (Pseudo-Elliptic filter, Broad-band E-plane filters and Hplane waveguide filters considering rounded corners). In order to enhance the capacities of the NN, we trained NN by the hybrids algorithms based on combining between back propagation (BP) algorithm and swarm intelligence algorithms (Social-Spider optimization SSO, spider monkey optimization SMO and Teaching–Learning-Based Optimization TLBO). To validate the training of neural networks using the proposed algorithms, we compared the results of convergence and modeling obtained with the results obtained using basic algorithms (SSO, SMO and TLBO) and also compared with population based algorithm, which is widely used in training NN namely particle swarm optimization (PSO). The results prove that the proposed hybrids algorithms have given better results.