, Volume 126, Issue 2, pp 315-350
Date: 05 May 2009

Smoothed analysis of condition numbers and complexity implications for linear programming

Rent the article at a discount

Rent now

* Final gross prices may vary according to local VAT.

Get Access


We perform a smoothed analysis of Renegar’s condition number for linear programming by analyzing the distribution of the distance to ill-posedness of a linear program subject to a slight Gaussian perturbation. In particular, we show that for every n-by-d matrix Ā, n-vector \({\bar{\varvec b}}\) , and d-vector \({\bar{\varvec c}}\) satisfying \({{||\bar{\bf A}, \bar{\varvec b}, \bar{\varvec c}||_F \leq 1}}\) and every σ ≤ 1, $$\mathop{\bf E}\limits_{\bf A,\varvec b,\varvec c }{{[\log C (\bf A,\varvec b,\varvec c)} = O (\log (nd/\sigma)),}$$ where A, b and c are Gaussian perturbations of Ā, \({\bar{\varvec b}}\) and \({\bar{\varvec c}}\) of variance σ 2 and C (A, b, c) is the condition number of the linear program defined by (A, b, c). From this bound, we obtain a smoothed analysis of interior point algorithms. By combining this with the smoothed analysis of finite termination of Spielman and Teng (Math. Prog. Ser. B, 2003), we show that the smoothed complexity of interior point algorithms for linear programming is O (n 3log(nd/σ)).

This work was supported in part by NSF Grants CCR-9875024, CCR-9972532, CCR-0112487, CCR-0325630, and CCF-0707522, and fellowships from the Alfred P. Sloan Foundation. Part of this work was done while John Dunagan and Daniel A. Spielman were at MIT.