Active Machine Learning for Adaptive and Efficient Sparse Near-Field Scanning: An Overview

Near-field scanning (NFS) is a commonly adopted technique in diagnosing electromagnetic interference (EMI) sources. However, NFS is a tedious and time-consuming process when facing a large number of scanning locations. An active machine learning method has been recently proposed to significantly accelerate NFS by adaptively selecting sparse samples. This paper gives an overview of the active learning method for sparse NFS. The basic principle of the active learning method is first explained. Then, applications of the technique on different NFS scenarios are discussed, including multi-frequency and multi-component measurements, time-domain measurements, and emission source microscopy measurements. Moreover, a stopping criterion to automatically terminate the sparse NFS process is presented. Potential applications of the active learning method in other electromagnetic compatibility problems are also discussed.