Guest Editorial

Abstract

Artificial intelligence (AI) and machine learning have experienced phenomenal success in the past decade in signal processing, image and speech recognition, robotics, autonomous systems, and more. This success is also coupled with the expanding applications of AI and machine learning in broad areas of science and engineering. The microwave community is among the earliest in exploring machine learning and artificial neural networks (ANN) for wireless and wireline electronic device, and circuit and system designs. Early works that appeared in the IEEE MTT-S International Microwave Symposium (IMS) are, for example, ANN for microstrip design (1993), ANN for microwave circuit analysis and optimization (1994), ANN for via interconnect (1996), and ANN for microwave computer-aided design (1996). The first time that IMS hosted a workshop on this topic (Applications of ANN for Microwave Design) was in IMS 1997, where the workshop was organized by Profs. K. C. Gupta and Michel Nakhla, and the invited speakers included R. L. Mahajan, K. C. Gupta, M. S. Nakhla, G. L. Creech, and Q. J. Zhang. Over the subsequent years, many researchers from around the world have contributed to the continuing development of ANN/machine learning techniques for microwave design. Several factors drove the development, one being the learning and generalization ability of ANN to learn computationally expensive microwave modeling/design problems offline, and to provide fast solutions to the problems during online design tasks. Another driving factor is the ability of ANN to learn microwave relationships directly from data even if analytical formulas for the targeted relationship are not available. From this angle, various ANN-based inverse modeling methods are developed, providing direct solutions to microwave design problems which otherwise would require expensive iterative process to solve. Further integrations between the microwave and machine learning disciplines led to knowledge-based approaches for modeling and design incorporating existing microwave-specific knowledges with the learning capabilities of ANNs, and technologies addressing cost of microwave data generation to perform effective machine learning for microwave applications.