LGRT 1685
Symmetry is ubiquitous in machine learning (ML) and many models produce invariant outputs when the input is transformed by different group actions. In the case when a data distribution possesses an intrinsic group-invariant structure, we are able to learn or approximate the underlying distribution more efficiently. In this talk, I will introduce the background of group symmetry and then present recent work on improved convergence. Finally, I will mention some future directions.