maspack.solvers

Class DantzigQPSolver

java.lang.Object

maspack.solvers.DantzigQPSolver

public class DantzigQPSolver
extends java.lang.Object

A dense QP (Quadratic Program) solver that that uses Dantzig's algorithm.

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	DantzigQPSolver.Status Described whether or not a solution was found.

Constructor Summary

Constructors

Constructor and Description
DantzigQPSolver()

Method Summary

Modifier and Type	Method and Description
DantzigQPSolver.Status	solve(VectorNd x, MatrixNd H, VectorNd f, MatrixNd A, VectorNd b) Solves a convex quadratic program with inequality constraints:
DantzigQPSolver.Status	solve(VectorNd x, MatrixNd H, VectorNd f, MatrixNd A, VectorNd b, MatrixNd Aeq, VectorNd beq) Solves a convex quadratic program with both equality and inequality constraints:

Methods inherited from class java.lang.Object

equals, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

DantzigQPSolver

public DantzigQPSolver()

Method Detail

solve

public DantzigQPSolver.Status solve(VectorNd x,
                                    MatrixNd H,
                                    VectorNd f,
                                    MatrixNd A,
                                    VectorNd b)

Solves a convex quadratic program with inequality constraints:

 
 min 1/2 x^T H x + f^T x,  A x >= b
 
 

using Dantzig's LCP pivoting algorithm.

Parameters:

x - computed minimum value

H - quadratic matrix term. Must be symmetric positive definite

f - linear term

A - inequality constraint matrix

b - inequality constraint offsets

Returns:

status value.

solve

public DantzigQPSolver.Status solve(VectorNd x,
                                    MatrixNd H,
                                    VectorNd f,
                                    MatrixNd A,
                                    VectorNd b,
                                    MatrixNd Aeq,
                                    VectorNd beq)

Solves a convex quadratic program with both equality and inequality constraints:

 
 min 1/2 x^T H x + f^T x, A x >= b, Aeq x = beq
 
 

using Dantzig's LCP pivoting algorithm. If there are no inequality constraints, then A should either have a row size of 0 or be specified as null.

Note that the number of inequality constraints (i.e., the row size of Aeq) should not exceed the size of H. Also, if the row size of Aeq equals the size of H, then only the equalites are solved for and the rest of the problem is ignored.

Parameters:

x - computed minimum value

H - quadratic matrix term. Must be symmetric positive definite

f - linear term

A - inequality constraint matrix (can be null)

b - inequality constraint offsets (can be null if A is null)

Aeq - equality constraint matrix

beq - equality constraint offsets

Returns:

status value.