A Newton-type method for sparse quadratic programming problems
We consider the quadratic programming problem min (1/2)x{sup T} Qx + q{sup T} x subject to Ax {<=} {alpha} with a n {times} n positive semidefinite matrix Q and a m {times} n matrix A. The method we want to propose is based on a special transformation of the above problem into an equivalent nonlinear system of equations. To solve this a damped Newton-type method will be suggested. In particular, the talk will deal with the following topics: (1) Solvability of all occurring Newton subproblems; (2) Global convergence; (3) Fast local convergence; (4) Decomposition of the subproblems with the aim that a sparse structure can be exploited efficiently; (5) Numerical experience, e.g. for sparse problems from shape preserving spline interpolation.
- OSTI ID:
- 36009
- Report Number(s):
- CONF-9408161--
- Country of Publication:
- United States
- Language:
- English
Similar Records
An Augmented Lagrangian Method for a Class of Inverse Quadratic Programming Problems
A Schur-complement method for sparse quadratic programming