Machine Learning and Data Reconstruction in Experimental Physics

During the last decade, research of machine learning (ML) and artificial intelligence (AI) has gone through an explosive evolution and made impacts across many domains of science and human lives. ML/AI models in computer vision can perform high quality analysis of image data from physics experiments, and models developed for natural language processing are applied to analyze sequential data. Graph models can be used to represent more general relations and thus powerful tools for analyzing data from multi-modal detector data in particle physics experiments. These models are promising not only for its performance on data reconstruction tasks, but also for automated optimization procedures that reduce much of human interventions needed for traditional, hand-engineered algorithms. Furthermore, the software eco-systems they are built on top of can exploit modern computing infrastructures such as High Performance Computing clusters and supported by interdisciplinary research communities across both academic and industrial research disciplines. In this talk, I introduce examples and challenges of ML/AI models that are used for data reconstruction tasks in experimental High Energy Physics with a focus on imaging detectors.