Talk Title: Factorizing Matrices with Missing Entries: Alternative Approaches

Speaker:  Prof. Charles Elkan

Abstract:

The known algorithms for approximate factorization of large matrices are so diverse that the multiple approaches have never been explained in one place,
and have never been fully compared experimentally. I will outline the available alternatives, focusing especially on experimental results, on methods for factorizing matrices with unknown entries, and on methods based on stochastic gradient descent.