Matrix Completion and the Netflix Prize

This video describes how the singular value decomposition (SVD) can be used for matrix completion and recommender systems.

These lectures follow Chapter 1 from: "Data-Driven Science and Engineering: Machine Learning, Dynamical Systems, and Control" by Brunton and Kutz

This video was produced at the University of Washington

[pedroguillermo]

Excellent explanation on SVD! Understood it very well and used it on my Bachelor's Thesis. Thank you! 😃

[alkiriiiic]

Man you carry my deep learning courses. Thank you a lot.

[PunmasterSTP]

Netflix prize? More like "No better videos exist before my eyes!" These lectures are incredible.

[RahulThakkar]

Thank you very much for this series!

I had a question about our earlier assumption about the value of m and n - I am assuming that it's not true anymore. Here m>n seems to be true. Does it change anything?
My understanding is that it only changes when we go for the economy option, we would take the whole of U, first nxn mini-square matrix of sigma and n columns of V.T. Are these assumptions correct? Does anything else change because of the m>n condition? Thanks!

[kwant]

How could we interpret the problem if matrix X was in transposed version? Eigen movie scores would appear?

[gaiuscaesar]

What is the difference between (normal SVD that contain USV^* that must provide 3 matrices) and (SVD that used by Simon Funk that contain 2 matrices)

[davidh]

I feel that I'm missing some intuition on the correlation matrices, does anyone know of good texts that discuss this in a relevant way? My Google-fu is failing me :(

[yannisstylianou]

It's really confusing how from our point of view, his writing should be mirrored, but its not.

[melihozcan]

@5:35 wish it worked in traffic as well :D