Overview
Save Big on Coursera Plus. 7,000+ courses at $160 off. Limited Time Only!
Explore robust regression techniques in this 47-minute lecture by Purushottam Kar from the International Centre for Theoretical Sciences. Delve into algorithms and optimization strategies, focusing on real-world applications such as recommendation systems and biometric identification. Learn about various solution strategies, including the Alternating Minimization Robust Regression (AM-RR) method, and understand why it works through detailed proofs. Examine generalized versions of AM-RR and their applications to non-toy problems. This talk, part of a discussion meeting on algorithms and optimization, offers insights into recent advances in learning algorithms, convex and nonconvex optimization, combinatorial optimization, and spectral algorithms.
Syllabus
Intro
A Recommendation System Problem
A Biometric Identification Problem
Robust Learning and Estimation - Application
A Toy Problem befitting this near-lunch Hour
Notation
Some Solution Strategies
An Alternate Viewpoint
AM-RR at work
Why AM-RR works?
The Proof
A Generalized AM-RR
Non-toy Problems for relaxed introspection
We do have some answers
Taught by
International Centre for Theoretical Sciences