An infeasible-interior-point algorithm for linear complementarity problems

We modify the algorithm of Zhang to obtain anO(n²L) infeasible-interior-point algorithm for monotone linear complementarity problems that has an asymptoticQ-subquadratic convergence rate. The algorithm requires the solution of at most two linear systems with the same coefficient matrix at each iteration.This research was supported by the Office of Scientific Computing, U.S. Department of Energy, under Contract W-31-109-Eng-38.     

AnO(nL) infeasible-interior-point algorithm for LCP with quadratic convergence

The Mizuno-Todd-Ye predictor-corrector algorithm for linear programming is extended for solving monotone linear complementarity problems from infeasible starting points. The proposed algorithm requires two matrix factorizations and at most three backsolves per iteration. Its computational complexity depends on the quality of the starting point. If the starting points are large enough, then the algorithm hasO(nL) iteration complexity. If a certain measure of feasibility at the starting point is small enough, then the algorithm has iteration complexity. At each iteration, both feasibility and optimality are reduced exactly at the same rate. The algorithm is quadratically convergent for problems having a strictly complementary solution, and therefore its asymptotic efficiency index is . A variant of the algorithm can be used to detect whether solutions with norm less than a given constant exist.This work was supported in part by the National Science Foundation under grant DMS-9305760.     

Superlinearly Convergent Infeasible-Interior-Point Algorithm for Degenerate LCP

A large-step infeasible path-following method is proposed for solving general linear complementarity problems with sufficient matrices. If the problem has a solution, the algorithm is superlinearly convergent from any positive starting points, even for degenerate problems. The algorithm generates points in a large neighborhood of the central path. Each iteration requires only one matrix factorization and at most three (asymptotically only two) backsolves. It has been recently proved that any sufficient matrix is a P _*()-matrix for some 0. The computational complexity of the algorithm depends on as well as on a feasibility measure of the starting point. If the starting point is feasible or close to being feasible, then the iteration complexity is . Otherwise, for arbitrary positive and large enough starting points, the iteration complexity is O((1 + )² nL). We note that, while computational complexity depends on , the algorithm itself does not.     

A unified approach to infeasible-interior-point algorithms via geometrical linear complementarity problems

There are many interior-point algorithms for LP (linear programming), QP (quadratic programming), and LCPs (linear complementarity problems). While the algebraic definitions of these problems are different from each other, we show that they are all of the same general form when we define the problems geometrically. We derive some basic properties related to such geometrical (monotone) LCPs and based on these properties, we propose and analyze a simple infeasible-interior-point algorithm for solving geometrical LCPs. The algorithm can solve any instance of the above classes without making any assumptions on the problem. It features global convergence, polynomial-time convergence if there is a solution that is smaller than the initial point, and quadratic convergence if there is a strictly complementary solution.This research was performed while the first author was visiting the Institute of Applied Mathematics and Statistics, Würzburg University as a Research Fellow of the Alexander von Humboldt Foundation.     

A superquadratic infeasible-interior-point method for linear complementarity problems

We consider a modification of a path-following infeasible-interior-point algorithm described by Wright. In the new algorithm, we attempt to improve each major iterate by reusing the coefficient matrix factors from the latest step. We show that the modified algorithm has similar theoretical global convergence properties to those of the earlier algorithm while its asymptotic convergence rate can be made superquadratic by an appropriate parameter choice. The work of this author was based on research supported by the Office of Scientific Computing, US Department of Energy, under Contract W-31-109-Eng-38. The work of this author was based on research supported in part by the US Department of Energy under Grant DE-FG02-93ER25171.     

A short proof of finiteness of Murty's principal pivoting algorithm

We give a short proof of the finiteness of Murty's principal pivoting algorithm for solving the linear complementarity problemy = Mz + q, y ^T z = 0,y 0,z 0 withP-matrixM.     

Predictor-corrector algorithm for solvingP *(κ)-matrix LCP from arbitrary positive starting points

A new predictor-corrector algorithm is proposed for solvingP _*(κ)-matrix linear complementarity problems. If the problem is solvable, then the algorithm converges from an arbitrary positive starting point (x ⁰,s ⁰). The computational complexity of the algorithm depends on the quality of the starting point. If the starting point is feasible or close to being feasible, it has -iteration complexity, whereρ ₀ is the ratio of the smallest and average coordinate ofX ⁰ s ⁰. With appropriate initialization, a modified version of the algorithm terminates in O((1+κ)²(n/ρ ₀)L) steps either by finding a solution or by determining that the problem has no solution in a predetermined, arbitrarily large, region. The algorithm is quadratically convergent for problems having a strictly complementary solution. We also propose an extension of a recent algorithm of Mizuno toP _*(κ)-matrix linear complementarity problems such that it can start from arbitrary positive points and has superlinear convergence without a strictly complementary condition. The work of this author was supported in part by NSF, Grant DMS 9305760 and by an Oberman fellowship from the University of Iowa Center for Advanced Studies.     

The global linear and local quadratic convergence of a non-interior continuation algorithm for the LCP

^** Email: zhenghaihuang{at}yahoo.com.cn; huangzhenghai{at}hotmail.com In this paper, we propose a non-interior continuation algorithmfor solving the P₀-matrix linear complementarity problem (LCP),which is conceptually simpler than most existing non-interiorcontinuation algorithms in the sense that the proposed algorithmonly needs to solve at most one linear system of equations ateach iteration. We show that the proposed algorithm is globallyconvergent under a common assumption. In particular, we showthat the proposed algorithm is globally linearly and locallyquadratically convergent under some assumptions which are weakerthan those required in many existing non-interior continuationalgorithms. It should be pointed out that the assumptions usedin our analysis of both global linear and local quadratic convergencedo not imply the uniqueness of the solution to the LCP concerned.To the best of our knowledge, such a convergence result hasnot been reported in the literature.     

On the Rate of Local Convergence of High-Order-Infeasible-Path-Following Algorithms for P*-Linear Complementarity Problems

A simple and unified analysis is provided on the rate of local convergence for a class of high-order-infeasible-path-following algorithms for the P_*-linear complementarity problem (P_*-LCP). It is shown that the rate of local convergence of a -order algorithm with a centering step is + 1 if there is a strictly complementary solution and ( + 1)/2 otherwise. For the -order algorithm without the centering step the corresponding rates are and /2, respectively. The algorithm without a centering step does not follow the fixed traditional central path. Instead, at each iteration, it follows a new analytic path connecting the current iterate with an optimal solution to generate the next iterate. An advantage of this algorithm is that it does not restrict iterates in a sequence of contracting neighborhoods of the central path.     

On the modulus algorithm for the linear complementarity problem

Concerning three subclasses of P-matrices the modulus algorithm and the projected successive overrelaxation (PSOR) method solving the linear complementarity problem are compared to each other with respect to convergence. It is shown that the modulus algorithm is convergent for all three subclasses whereas the convergence of the PSOR method is only guaranteed for two of them.     

A primal—dual infeasible-interior-point algorithm for linear programming

As in many primal—dual interior-point algorithms, a primal—dual infeasible-interior-point algorithm chooses a new point along the Newton direction towards a point on the central trajectory, but it does not confine the iterates within the feasible region. This paper proposes a step length rule with which the algorithm takes large distinct step lengths in the primal and dual spaces and enjoys the global convergence.Part of this research was done when M. Kojima and S. Mizuno visited at the IBM Almaden Research Center. Partial support from the Office of Naval Research under Contract N00014-91-C-0026 is acknowledged.Supported by Grant-in-Aids for Co-operative Research (03832017) of The Japan Ministry of Education, Science and Culture.Supported by Grant-in-Aids for Encouragement of Young Scientist (03740125) and Co-operative Research (03832017) of The Japan Ministry of Education, Science and Culture.     

A note on solvability of a class of linear complementarity problems

We give a characterization of unique solvability of an infinite family of linear complementarity problems of a special form by means of a finite subset of this family.     

A superlinearly convergent predictor-corrector method for degenerate LCP in a wide neighborhood of the central path with $$O(\sqrt{n}L)$$-iteration complexity

We correct an error in Algorithm 2 from [1]. The online version of the original article can be found at     

Global Linear and Quadratic One-step Smoothing Newton Method for P 0-LCP

We propose a new smoothing Newton method for solving the P ₀-matrix linear complementarity problem (P ₀-LCP) based on CHKS smoothing function. Our algorithm solves only one linear system of equations and performs only one line search per iteration. It is shown to converge to a P ₀-LCP solution globally linearly and locally quadratically without the strict complementarity assumption at the solution. To the best of author's knowledge, this is the first one-step smoothing Newton method to possess both global linear and local quadratic convergence. Preliminary numerical results indicate that the proposed algorithm is promising.     

A Non-Interior Continuation Algorithm for the P0 or P* LCP with Strong Global and Local Convergence Properties

We propose a non-interior continuation algorithm for the solution of the linear complementarity problem (LCP) with a P₀ matrix. The proposed algorithm differentiates itself from the current continuation algorithms by combining good global convergence properties with good local convergence properties under unified conditions. Specifically, it is shown that the proposed algorithm is globally convergent under an assumption which may be satisfied even if the solution set of the LCP is unbounded. Moreover, the algorithm is globally linearly and locally superlinearly convergent under a nonsingularity assumption. If the matrix in the LCP is a P_* matrix, then the above results can be strengthened to include global linear and local quadratic convergence under a strict complementary condition without the nonsingularity assumption.     

Existence and uniqueness of solutions for the generalized linear complementarity problem

Cottle and Dantzig (Ref. 1) showed that the generalized linear complementarity problem has a solution for anyqR ^m ifM is a vertical blockP-matrix of type (m ₁,...,m _n ). They also extended known characterizations of squareP-matrices to vertical blockP-matrices.Here we show, using a technique similar to Murty's (Ref. 2), that there exists a unique solution for anyqR ^m if and only ifM is a vertical blockP-matrix of type (m ₁,...,m _n ). To obtain this characterization, we employ a generalization of Tucker's theorem (Ref. 3) and a generalization of a theorem initially introduced by Gale and Nikaido (Ref. 4).     

A superlinearly convergent predictor-corrector method for degenerate LCP in a wide neighborhood of the central path with -iteration complexity

An interior point method for monotone linear complementarity problems acting in a wide neighborhood of the central path is presented. The method has -iteration complexity and is superlinearly convergent even when the problem does not possess a strictly complementary solution. Mathematics Subject Classification (2000): 49M15, 65K05, 90C33 Work supported by the National Science Foundation under Grant No. 0139701. An erratum to this article is available at.     

On quadratic andO\left( {\sqrt {nL} } \right) convergence of a predictor—corrector algorithm for LCP

Recently several new results have been developed for the asymptotic (local) convergence of polynomial-time interior-point algorithms. It has been shown that the predictor—corrector algorithm for linear programming (LP) exhibits asymptotic quadratic convergence of the primal—dual gap to zero, without any assumptions concerning nondegeneracy, or the convergence of the iteration sequence. In this paper we prove a similar result for the monotone linear complementarity problem (LCP), assuming only that a strictly complementary solution exists. We also show by example that the existence of a strictly complementarity solution appears to be necessary to achieve superlinear convergence for the algorithm.Research supported in part by NSF Grants DDM-8922636 and DDM-9207347, and an Interdisciplinary Research Grant of the University of Iowa, Iowa Center for Advanced Studies.     

Strict Feasibility Conditions in Nonlinear Complementarity Problems

Strict feasibility plays an important role in the development of the theoryand algorithms of complementarity problems. In this paper, we establishsufficient conditions to ensure strict feasibility of a nonlinearcomplementarity problem. Our analysis method, based on a newly introducedconcept of -exceptional sequence, can be viewed as a unified approachfor proving the existence of a strictly feasible point. Some equivalentconditions of strict feasibility are also developed for certaincomplementarity problems. In particular, we show that aP_*-complementarity problem is strictly feasible if and only ifits solution set is nonempty and bounded.     

Solving more linear complementarity problems with Murty's bard-type algorithm

The linear complementarity problem (M|q) is to findw andz inR ⁿ such thatw–Mz=q,w0,z0,w ^t z=0, givenM inR ^n×n andq in . Murty's Bard-type algorithm for solving LCP is modeled as a digraph.Murty's original convergence proof considered allq inR ⁿ andM inR ^n×n , aP-matrix. We show how to solve more LCP's by restricting the set ofq vectors and enlarging the class ofM matrices beyondP-matrices. The effect is that the graph contains an embedded graph of the type considered by Stickney and Watson wheneverM is a matrix containing a principal submatrix which is aP-matrix. Examples are presented which show what can happen when the hypotheses are further weakened.