Applying neural network for RF and electrical semiconductor to get an optimal structure

Tendency and requirement to transform and process information quickly forced us todesign and make circuits that can work in high frequency.. At this time there are differenttypes of high frequency devices for which their models for CAD are necessary. As far as I knowin the latest version of Spice only a class of five types of the MESFET model is available.Because of Advanced ANN training creates accurate & general model functions. In thispaper, a method is suggested for modeling RF and electronic semiconductors by exclusiveneural networks or by corrective neural networks working attached to a modified analyticmodel. And finding an optimal structure which will be applicable for HEMT, MESFET andother types of FETs and DGMESFET . An accuracy of the proposed modification of theanalytic model is assessed by extracting model parameters of the AlGaAs/InGaAs/GaAspHEMT. An accuracy of process with neural networks is generally assessed by extracting their parameters in static and dynamic domains.