Reverse modeling of microwave circuits with bidirectional neural network models
M. Vai, Shuichi Wu, Bin Li, Sudhanand Prasad · IEEE Transactions on Microwave Theory and Techniques · 1998
Neural networks have been been developed into an alternative modeling approach for the microwave circuit-design process. In this paper, we describe a neural network-based microwave circuit-design approach that implements the solution-searching optimization routine by a modified neural network learning process. Both the development of a microwave circuit model and the searching of a design solution can thus take advantage of a hardware neural network processor, which is significantly faster than a software simulation. In addition, a systematic simulation-based approach to convert conventional circuit models into neural network models for this design process is described. The development of a heterojunction bipolar transistor (HBT) amplifier model and its applications are demonstrated.