Preprocessing and Regularization for Degenerate Semidefinite Programs

Conference paper

DOI: 10.1007/978-1-4614-7621-4_12

Part of the Springer Proceedings in Mathematics & Statistics book series (PROMS, volume 50)
Cite this paper as:
Cheung YL., Schurr S., Wolkowicz H. (2013) Preprocessing and Regularization for Degenerate Semidefinite Programs. In: Bailey D. et al. (eds) Computational and Analytical Mathematics. Springer Proceedings in Mathematics & Statistics, vol 50. Springer, New York, NY

Abstract

This paper presents a backward stable preprocessing technique for (nearly) ill-posed semidefinite programming, SDP, problems, i.e., programs for which the Slater constraint qualification (SCQ), the existence of strictly feasible points, (nearly) fails. Current popular algorithms for semidefinite programming rely on primal-dual interior-point, p-d i-p, methods. These algorithms require the SCQ for both the primal and dual problems. This assumption guarantees the existence of Lagrange multipliers, well-posedness of the problem, and stability of algorithms. However, there are many instances of SDPs where the SCQ fails or nearly fails. Our backward stable preprocessing technique is based on applying the Borwein–Wolkowicz facial reduction process to find a finite number, k, of rank-revealing orthogonal rotations of the problem. After an appropriate truncation, this results in a smaller, well-posed, nearby problem that satisfies the Robinson constraint qualification, and one that can be solved by standard SDP solvers. The case k = 1 is of particular interest and is characterized by strict complementarity of an auxiliary problem.

Key words

Backward stability Degeneracy Preprocessing Semidefinite programming Strong duality 

Mathematics Subject Classifications (2010)

Primary, 49K40 90C22 Secondary, 65K10 90C25 90C46 

Copyright information

© Springer Science+Business Media New York 2013

Authors and Affiliations

  • Yuen-Lam Cheung
    • 1
  • Simon Schurr
    • 1
  • Henry Wolkowicz
    • 1
  1. 1.Department of Combinatorics and OptimizationUniversity of WaterlooWaterlooCanada

Personalised recommendations