Course: STAT 37788=CMSC 35430
Title: Machine Learning on Graphs, Groups, and Manifolds
Instructor(s): Risi Kondor
Class Schedule: Sec 1: MW 1:30 PM-2:50 PM in Ryerson 176
Description: In many domains, including applications of machine learning to scientific problems, social phenomena and computer vision/graphics, the data that learning algorithms operate on naturally lives on structured objects such as graphs or low dimensional manifolds. There are many connections between these cases; further, since groups capture symmetries, there are also natural connections to the theory of learning on groups and group equiviariant algorithms. This course provides a mathematical introduction to these topics both in the context of kernel based learning and neural networks. Specific topics covered include graph kernels, manifold learning, graph wavelets, graph neural networks, permutation equivariant learning, rotational equivariant networks for scientific applications and imaging, gauge equivariant networks and steerable nets.