AI/ML for Microwave Modeling and Optimization

Accurate modeling and optimization of microwave components and subsystems are essential for microwave design. This talk starts with a brief overview of AI and ML methodologies in modeling and optimization of passive/active microwave devices and circuits, followed by descriptions of several important directions of development. Automated model generation algorithms facilitate adaptive data generation for machine learning, and feature neural network structural adaptation. Knowledge-based neural network incorporates prior microwave knowledge in the form of microwave equivalent and empirical models into neural network structures, ensuring ML’s reliability when training data is limited, and enhancing ML’s extrapolation capability. AI/ML substantially speeds-up electromagnetic-based microwave modeling and optimization. Applications in electromagnetic based modeling and optimization of microwave filters and passive/active devices will be presented.