#include <Jacobian.h>

Inheritance diagram for Minotaur::Jacobian:

Public Member Functions
	Jacobian ()
	Default constructor.

	Jacobian (const std::vector< ConstraintPtr > &cons, const UInt n)
	Constructor using constraints.

virtual	~Jacobian ()
	Destroy.

virtual UInt	getNumNz ()
	Return the number of nonzeros in the Jacobian. More...

virtual void	fillRowColIndices (UInt iRow, UInt jCol)

virtual void	fillRowColValues (const double x, double values, int *error)

virtual void	fillColRowIndices (UInt , UInt )
	Fill indices, column wise. More...

virtual void	fillColRowValues (const double , double , int *)
	Fill values, column wise. More...

void	write (std::ostream &out) const

Detailed Description

This class is used for the Jacobian of a Problem. When a problem has constraints of the form:

$g_i(x) \leq 0$

The Jacobian at a point $x^*$ is a matrix whose $i-$ th row is the gradient, $\nabla f_i(x^*)$ .

This class also provides the sparsity pattern of the non-zeros in the Jacobian. By default, we save the Jacobian constraint-wise and then variable-wise. Thus, if we a have the constraints:

x_0^2 + x1^3 \leq 3 x_0 + x_2 + x_3 \leq 4 x_1 + x_3 \leq 4 x_3^3 \leq 8

We will save the indices of nonzeros in a 2-d matrix, rows corresponding to constraints and columns corresponding to variables. The matrix will have entries: 0 1 0 2 3 1 3 3

The above scheme makes it easy to fill nonzeros in Constraint-Ordered manner, but more expensive to fill in Variable-Ordered manner. So we also use another 2-d matrix as above, but the matrix now looks like: 0 0 1 1 1 2 2 3

Member Function Documentation

◆ fillColRowIndices()

virtual void Minotaur::Jacobian::fillColRowIndices	(	UInt *	,
		UInt *
	)

inlinevirtual

Fill indices, column wise.

Reimplemented in MINOTAUR_AMPL::AMPLJacobian.

◆ fillColRowValues()

virtual void Minotaur::Jacobian::fillColRowValues	(	const double *	,
		double *	,
		int *
	)

inlinevirtual

Fill values, column wise.

Reimplemented in MINOTAUR_AMPL::AMPLJacobian.

◆ fillRowColIndices()

void Jacobian::fillRowColIndices	(	UInt *	iRow,
		UInt *	jCol
	)

virtual

Given arrays iRow and jCol, fill in the row and column index of each non-zero in the jacobian. e.g.

$\begin{eqnarray*} x_1x_3 + x_1^2x_4 &=& 2 \\ x_0 + x_5 &=& 0 \end{eqnarray*}$

then, iRow = [1 0 0 0 1], jCol = [0 1 3 4 5]. These indices are arranged first in the order of increasing jCol and then increasing iRow.

Reimplemented in MINOTAUR_AMPL::AMPLJacobian.

◆ fillRowColValues()

void Jacobian::fillRowColValues	(	const double *	x,
		double *	values,
		int *	error
	)

virtual

Given arrays iRow and jCol as above, fill in the row and column index of each non-zero in the jacobian. For the above example, when $x =
(-1, 1, 1, 1, 0)$ , then values = [3 -1 1 1 1].

Reimplemented in MINOTAUR_AMPL::AMPLJacobian.

◆ getNumNz()

UInt Jacobian::getNumNz ( )

virtual

Return the number of nonzeros in the Jacobian.

Reimplemented in MINOTAUR_AMPL::AMPLJacobian.

The documentation for this class was generated from the following files:

/home/amahajan/tmp/minotaur-test/src/base/Jacobian.h
/home/amahajan/tmp/minotaur-test/src/base/Jacobian.cpp

Public Member Functions

Detailed Description

Member Function Documentation

◆ fillColRowIndices()

◆ fillColRowValues()

◆ fillRowColIndices()

◆ fillRowColValues()

◆ getNumNz()