----------------------------------------------------------------------------------------------------

     Department of Systems Engineering and Engineering Management

                    The Chinese University of Hong Kong

----------------------------------------------------------------------------------------------------

Date: Friday, September 25, 2026, 4:30pm to 5:30pm HKT

Venue: ERB 909, The Chinese University of Hong Kong

Title:  Learning Sparse Representations with Symmetries

Speaker: Prof. Yong Sheng Soh, National University of Singapore

Abstract:

Symmetries and invariances feature in wide range of learning problems.  As a prominent example, translation invariance underpins the construction of convolutional neural networks used in vision tasks.  In this talk, we will consider the task of learning data representations that obey some pre-specified notion of symmetry (e.g. translations or rotations).  For concreteness, we focus on the specific task of learning sparse representations -- or, equivalently, learning regularisers -- for a dataset.  We prescribe an end-to-end recipe that can be instantiated for compact topological groups.  Important ideas that feature are (i) the representation theory of (compact) groups, which allows us to express these symmetries in terms of irreducible representations, (ii) harmonic analysis, which allows us to leverage powerful ideas from Fourier analysis to facilitate computation, and (iii) convex geometry, which allows us to formulate these representation learning tasks as instances of structured conic programs.

Biography:

Yong Sheng Soh is an Assistant Professor at the Department of Mathematics, National University of Singapore.  He is broadly interested in mathematical optimization, with a focus on problems arising from data analysis.  He obtained his PhD in Applied and Computational Mathematics at the California Institute of Technology.  Prior to joining NUS, he was a Research Scientist at the Institute for High Performance Computing at the Agency for Science, Technology and Research (A*STAR), Singapore.

Everyone is welcome to attend the talk!

SEEM-5201 Website: http://seminar.se.cuhk.edu.hk