Mini-course on Representation Theoretical Methods in Machine Learning

Risi Kondor, Gatsby Unit


Wednesday, October 1, 1.20-2.20 [slides]
Friday, October 3, 4.30-5.30 [slides]
Wednesday, October 8, 1.20-2.20 [slides]
Friday, October 10, 4.30-5.30 [slides]
Friday, October 17, 4.30-5.30 [slides]
Friday, October 24, 4.30-5.30 [slides]
Friday, October 31, 4.30-5.30 (see previous lecture)

Location: 4th floor seminar room, Alexandra House


This series of four lectures is a short introduction to non-commutative harmonic analysis and its applications in machine learning and statistics. No prior knowledge of group theory or representation theory is assumed.

Topics covered will include:
- the role of groups in mathematics, physics and statistics;
- introduction to representation theory;
- harmonic analysis -- what is it really about?
- the representation theory of the symmetric group;
- the representation theory of the rotation groups;
- fast Fourier transforms and Clausen's algorithm;
- invariants: power spectrum, bispectrum and skew spectrum;
- optimization algorithms in the Fourier domain;
- permutation polytopes, QAPs, traveling salesman, etc.;
- random walks on groups;
- applications in multi-object tracking;
- applications in graph computations.

Please let me know if you are interested in attending, if you have not already done so. Slides and references will be posted on this page.