Course Identification

1. Basic knowledge in deep learning (a course in deep learning; implemented and trained a deep learning model).
2. A basic course in algebraic structures (familiarity with the concept of a group/ group action).
3. Undergraduate courses in calculus and linear algebra (basic topics).

Summary and outcomes:

This is a seminar course that will cover recent advances in geometric and algebraic methods in deep learning. Introduction will be given by course organizers. We will cover recent advances that deal with images, volumetric images, sets, point clouds, graphs, spheres, and manifolds. 

Every participants is required to read all papers and write 1-2 paragraph summary for each, and present one of the papers.

Upon successful completion of this course students will:

Get familiar with the current frontier of applying deep learning to irregular data with different notions of equivalence. For example, two graphs are equivalent if they are the same up to relabeling the nodes. Therefore, when learning graphs we would like to use networks that are invariant/equivariant to the relabeling operation.