-
Title invited talk:
Low-Rank Matrix Completion (LRMC) using Nuclear Norm (NN) with Facial
Reduction (FR); Applications to Robust Principal Component Analysis
Henry Wolkowicz, University of Waterloo, Waterloo, ON, Canada. Contact:
hwolkowicz@uwaterloo.ca
Minimization of NN is often used as a convex relaxation for solving LRMC
problems. This can then be solved using semidefinite programming (SDP).
The SDP and its dual are regular. FR has been successful in regularizing
degenerate problems. Here we take advantage of the structure at
optimality for the NN minimization and show that even though strict
feasibility holds, the FR framework can be successfully applied to
obtain a proper face that contains the optimal set and even improves on
the NN relaxation. We include numerical tests for both exact and noisy
cases.
-
Tutorial Session 90 minutes Title:
- Facial Reduction in Cone Optimization with Applications to Matrix
Completions
Add to Itinerary
November 5, 2018, 4:30 PM - 6:00 PM 122B, North Bldg
Authors
Henry Wolkowicz, University of Waterloo, Waterloo, ON, Canada.
Contact: hwolkowicz@uwaterloo.ca
Abstract
Strictly feasibility is at the heart of convex optimization. This is
needed for optimality conditions, stability, and algorithmic
development. New optimization modelling techniques and convex
relaxations for hard nonconvex problems have shown that the loss of
strict feasibility is a much more pronounced phenomenon than previously
realized. These new developments suggest a reappraisal. We describe the
various reasons for the loss of strict feasibility, whether due to poor
modelling choices or (more interestingly) rich underlying structure, and
describe ways to cope with it and, in particular, "take advantage of
it".