Design of Waveguide Matched Load Based on Multilayer Perceptron Neural Network
Renjie Zhu · 2007
Rectangular waveguide terminal matched load is designed based on multilayer perceptron neural network (MLPNN) that is trained by error back propagation (BP) method. One of improvements is that the structure of MLPNN can be adjusted automatically according to complicacy of modeled system. Simultaneously, objective function of MLPNN is modified based on information entropy theory. Training and testing sample are sorted randomly to make them more effective. As a result, H-plane T-kind terminal matched load of rectangular waveguide is designed successfully.