CDS Professor Presents at the 24th International Conference on Artificial Intelligence and Statistics
Last week, Kyle Cranmer, CDS professor of data science and professor of physics at the NYU College of Arts & Science gave a talk at AISTATS 2021, the 24th International Conference on Artificial Intelligence and Statistics. Kyle was joined by Emmanuel Candès, the Barnum-Simons chair in mathematics and statistics and professor of mathematics, of statistics, and of electrical engineering at Stanford University as well as Bin Yu, professor of of statistics and electrical engineering and computer sciences at UC Berkeley.
The conference was held virtually from Tuesday, April 13 to Thursday, April 15, 2021. Presented by the Society of AI and Statistics, the seminar is an interdisciplinary joining of researchers from areas such as artificial intelligence, machine learning, statistics, and other related areas, and has been active since 1985.
Kyle’s talk focused on how physical sciences are replete with high-fidelity simulators i.e. computational manifestation of causal, mechanistic models. Though these simulators provide the highest-fidelity physical models, they are not particularly convenient for inferring properties of the model from data. In this talk, Kyle described how machine learning and probabilistic programming techniques are being applied to these challenging problems. His presentation closed with examples of how these methods impact the study of particle physics at Large Hadron Collider (i.e. LHC, the world’s largest and most powerful particle accelerator) and in fields such as astrophysics, neuroscience, and public health. His presentation is largely based on “The frontier of simulation-based inference”, a paper he co-authored with former CDS-Moore Sloan fellows Johann Brehmer, currently deep learning research engineer at Qualcomm with Dr. Max Welling’s group and Gilles Louppe, associate professor in artificial intelligence and deep learning at the University of Liège in Belgium.
This lecture is essentially an expansion on research from Kyle’s NeurIPS 2016 Keynote presentation “Machine Learning & Likelihood Free Inference in Particle Physics” which we covered in our 2017 blog post “From Micro to Macro: Kyle Cranmer talks Machine Learning for the Natural Sciences”. Themes and elements related to this research are explored in the “Machine Learning for Physical Sciences” workshop series that Kyle co-founded.
By Ashley C. McDonald