[memexy]

Are there any good references for this approach (via generalized Fourier transforms)?

[bollu]

There is a thesis by Kondor: "Group theoretical methods for machine learning" - https://people.cs.uchicago.edu/~risi/papers/KondorThesis.pdf which provides links to the associated literature.

The documentation of his library (SnOB) is great: http://www.gatsby.ucl.ac.uk/~risi/SnOB/SnOB/SnOB.pdf

introductory slides on representation theory in machine learning: http://www.gatsby.ucl.ac.uk/~risi/courses/mini08/symmetric.p...

Full mini course: http://www.gatsby.ucl.ac.uk/~risi/courses/mini08/mini08.html

[memexy]

Thanks. These all look very readable.

[kcoul]

I wouldn’t call this 3blue1brown video a reference, but until I watched this (coming from an audio background), I didn’t quite grasp that a continuous stencil of any complex silhouette could be described in terms of a set of contributing vectors. Helped click that there is a perfect visual analogy to the deconstruction of complex audio to sine waves.

https://youtu.be/r6sGWTCMz2k

[sdenton4]

Persi Diaconis' book is excellent. You come away with the proof that it takes seven shuffles to fully randomize a deck of cards...

https://www.amazon.com/Group-Representations-Probability-Sta...

[pathsjs]

Yes: Folland, A course in abstract harmonic analysis, CRC press