Neural Space Mapping as a Pioneering Artificial Intelligence approach to Microwave Modeling and Design

Artificial Neural Networks (ANN), inspired from rudimentary biological neural networks, are nowadays a well-stablished, powerful, and general-purpose fundamental technique of Artificial Intelligence (AI). On the other hand, Space Mapping (SM), inspired in engineering intuition, intelligently exploits the computational efficiency of inaccurate simplified physics-based coarse models to optimally design highly accurate but computationally expensive fine models in a very efficient manner. In this presentation, a tribute is paid to Prof. John W. Bandler who pioneered the efficient combination of ANN and SM to formulate EM-based modeling and design optimization algorithms. In particular, Neural Space Mapping (NSM) and Neural Inverse Space Mapping (NISM) are briefly reviewed for EM-based design optimization, statistical analysis, and yield optimization of microwave circuits. Speculation on some future developments to capitalize the current full capacity of other artificial intelligence and machine learning techniques in combination with space mapping is also envisioned.