Review on Physics-Driven Learning for Inverse Problems Published in Nature Reviews Physics

Nature Reviews Physics recently published a review paper on ”Physics-driven learning for inverse problems in quantum chromodynamics”, from DEEP-IN working group led by Lingxiao Wang (Research Scientist, iTHEMS), including Tetsuo Hatsuda (Program Director, iTHEMS). The authors explore how combining deep learning techniques with physics-driven designs significantly improves the extraction of accurate physical properties from complex QCD phenomena. Highlighting applications such as lattice QCD calculations and studies of hadron interactions, neutron stars, and heavy-ion collisions, the review emphasizes the benefits of embedding physical priors into machine learning models and suggests broader potentials to general physics problems.