The session welcomes all papers that are related to learning approach to electromagnetic inverse problems. Both types of inverse problems are within the scope of the session, i.e., parameter identification and inverse design. The former usually presents unique ground truth, such as in inverse scattering, localization, and CT imaging. The latter usually admits multiple solutions, such as design of photonic crystal, antenna design, and array synthesis. The session aims at presenting frontier research in the theories, computations, and applications of machine learning for solving electromagnetic inverse problems. All types of learning algorithms are welcomed: supervised and semi-supervised learning, unsupervised learning, and reinforcement learning. Uncertainty quantification (UQ) analysis is also an important topic of this session. Applications include but not limited to microwave imaging, through-wall-imaging, non-line-of-sight imaging, subsurface detection, electric impedance or capacitance tomography, biomedical imaging, near-field optical imaging, non-destructive testing/evaluation, and remote sensing.