| Grant number: | 11/04289-0 |
| Support type: | Scholarships in Brazil - Master |
| Effective date (Start): | August 01, 2011 |
| Effective date (End): | March 31, 2013 |
| Field of knowledge: | Physical Sciences and Mathematics - Mathematics |
| Principal Investigator: | Marina Andretta |
| Grantee: | Jeinny Maria Peralta Polo |
| Home Institution: | Instituto de Ciências Matemáticas e de Computação (ICMC). Universidade de São Paulo (USP). São Carlos , SP, Brazil |
Abstract The linearly constrained minimization problem is important, not only for the problem itself, that appears in several areas, but because it is used as a subproblem of more general problems, such as the nonlinear programming problems. In [M. Andretta, E. G. Birgin e J. M. Martínez. Partial spectral projected gradient method with active-set strategy for linearly constrained optimization. Numerical Algorithms 53, pp. 23-52, 2010] was presented an efficient method for solving small and medium scaled linearly constrained minimization problems called GENLIN. To implement a similar method to solve large scale problems, it is necessary to have an efficient method to solve sparse projection problems onto linear constraints. The problem of projecting a point onto a set of linear constraints can be written as a convex quadratic programming problem. the goal of this work is to study and implement sparse methods to solve convex quadratic programming problems, in particular the classical Moré-Toraldo method, compare the performance of the methods when solving the projection onto linear constraints problem and adapt GENLIN to use sparse projections (when it is convenient). (AU) | |