dsdpcone.h File Reference

Detailed Description

The public interface between the cones and the solver.

Definition in file dsdpcone.h.

Go to the source code of this file.

Typedefs

typedef struct DSDPCone_C DSDPCone

This object holds the data of a SDP, LP, or other cone. Its structure is opaque to the DSDP Solver, but it must implement the interface below and provide a structure of function pointers.

Functions

int DSDPConeANorm2 (DSDPCone, DSDPVec)

Add square of 2-norm of data correponding to each variable y.

int DSDPConeComputeHessian (DSDPCone, double, DSDPSchurMat, DSDPVec, DSDPVec)

Compute Hessian and gradient of barrier function.

int DSDPConeComputeLogSDeterminant (DSDPCone, double *, double *)

Evaluate logrithmic barrier function.

int DSDPConeComputeMaxStepLength (DSDPCone, DSDPVec, DSDPDualFactorMatrix, double *)

Determine distance to the edge of the cone.

int DSDPConeComputeRHS (DSDPCone, double, DSDPVec, DSDPVec, DSDPVec)

Compute gradient of barrier function.

int DSDPConeComputeS (DSDPCone, DSDPVec, DSDPDualFactorMatrix, DSDPTruth *)

Given y, compute S and determine whether its in the cone.

int DSDPConeComputeX (DSDPCone, double, DSDPVec, DSDPVec, DSDPVec, double *)

Given y,dy, and mu, construct X and add its inner product with the data and S.

int DSDPConeDestroy (DSDPCone *)

Free the internal memory of the cone.

int DSDPConeGetDimension (DSDPCone, double *)

Provide the dimension of the cone.

int DSDPConeInitialize (DSDPCone *)

Initialize the pointers to 0.

int DSDPConeInvertS (DSDPCone)

Invert the dual matrix S.

int DSDPConeMonitor (DSDPCone, int)

Do anything at in the cone at each iteration.

int DSDPConeMultiplyAdd (DSDPCone, double, DSDPVec, DSDPVec, DSDPVec)

Multiply Hessian by a vector and add the result.

int DSDPConeSetData (DSDPCone *, struct DSDPCone_Ops *, void *)

Initialize the pointers to 0.

int DSDPConeSetUp (DSDPCone, DSDPVec)

Factor the data and allocate data structures.

int DSDPConeSetUp2 (DSDPCone, DSDPVec, DSDPSchurMat)

Factor the data and allocate data structures.

int DSDPConeSetXMaker (DSDPCone, double, DSDPVec, DSDPVec)

Pass information needed to construct X.

int DSDPConeSparsityInSchurMat (DSDPCone, int, int[], int)

Identify sparsity pattern in a row of the Hessian term.

int DSDPConeView (DSDPCone)

View contents of the cone.

int DSDPGetConeName (DSDPCone, char *, int)

Get name of the cone.

Typedef Documentation

struct DSDPCone DSDPCone

This object holds the data of a SDP, LP, or other cone. Its structure is opaque to the DSDP Solver, but it must implement the interface below and provide a structure of function pointers.

Definition at line 27 of file dsdpcone.h.

Function Documentation

int DSDPConeANorm2	(	DSDPCone	K,
		DSDPVec	anorm2
	)

Add square of 2-norm of data correponding to each variable y.

Parameters:

	K	the cone
	anorm2	norm of constraint data for each varibles

See also:: DSDPBlockANorm2

Definition at line 168 of file dsdpcone.c.

Referenced by DSDPComputeANorm2().

int DSDPConeComputeHessian	(	DSDPCone	K,
		double	mu,
		DSDPSchurMat	M,
		DSDPVec	vrhs1,
		DSDPVec	vrhs2
	)

Compute Hessian and gradient of barrier function.

Parameters:

	K	the cone
	mu	barrier parameter
	M	Schur matrix
	vrhs1	objective gradient
	vrhs2	barrier gradient

This routine assumes that the dual matrix has already been factored and inverted.

See also:: SDPConeComputeHessian()

Definition at line 92 of file dsdpcone.c.

Referenced by DSDPComputeHessian().

int DSDPConeComputeLogSDeterminant	(	DSDPCone	K,
		double *	logdetobj,
		double *	logdet
	)

Evaluate logrithmic barrier function.

Parameters:

	K	the cone
	logdetobj	used term.
	logdet	logarithmic barrier of cone Assumes S is in cone.

Definition at line 403 of file dsdpcone.c.

Referenced by DSDPComputeLogSDeterminant().

int DSDPConeComputeMaxStepLength	(	DSDPCone	K,
		DSDPVec	DY,
		DSDPDualFactorMatrix	flag,
		double *	maxsteplength
	)

Determine distance to the edge of the cone.

Parameters:

	K	the cone
	DY	step direction
	flag	identifies which of two S matrix structures should be used.
	maxsteplength	distance to the edge of the cone.

Definition at line 288 of file dsdpcone.c.

Referenced by DSDPComputeMaxStepLength().

int DSDPConeComputeRHS	(	DSDPCone	K,
		double	mu,
		DSDPVec	vrow,
		DSDPVec	rhs1,
		DSDPVec	rhs2
	)

Compute gradient of barrier function.

Parameters:

	K	the cone
	mu	barrier parameter
	vrow	scaling for each element in the gradient.
	rhs1	objective gradient
	rhs2	barrier gradient

This routine assumes that the dual matrix has already been factored and inverted. Define rhs2 += mu * vrow .* A(S^{-1})

See also:: SDPConeComputeRHS()

Definition at line 147 of file dsdpcone.c.

Referenced by DSDPComputeG().

int DSDPConeComputeS	(	DSDPCone	K,
		DSDPVec	Y,
		DSDPDualFactorMatrix	flag,
		DSDPTruth *	ispsdefinite
	)

Given y, compute S and determine whether its in the cone.

Parameters:

	K	the cone
	Y	solution
	flag	identifies which of two S matrix structures should be used.
	ispsdefinite	true if S is positive definite or an element of the cone.

Definition at line 242 of file dsdpcone.c.

Referenced by DSDPComputeSS().

int DSDPConeComputeX	(	DSDPCone	K,
		double	mu,
		DSDPVec	y,
		DSDPVec	dy,
		DSDPVec	AX,
		double *	tracexs
	)

Given y,dy, and mu, construct X and add its inner product with the data and S.

Parameters:

	K	the cone
	mu	barrier parameter
	y	solution
	dy	step direction
	AX	add the inner product of the data with X
	tracexs	inner product of X and S.

See also:: SDPConeComputeXX()

Definition at line 216 of file dsdpcone.c.

Referenced by DSDPComputeXVariables().

int DSDPConeDestroy ( DSDPCone * K )

Free the internal memory of the cone.

Parameters:

K

the cone

Definition at line 64 of file dsdpcone.c.

Referenced by DSDPDestroyCones().

int DSDPConeGetDimension	(	DSDPCone	K,
		double *	n
	)

Provide the dimension of the cone.

Parameters:

	K	the cone
	n	conic dimension (an integer value)

Definition at line 312 of file dsdpcone.c.

Referenced by DSDPGetConicDimension().

int DSDPConeInitialize ( DSDPCone * K )

Initialize the pointers to 0.

Parameters:

K

the cone

Definition at line 495 of file dsdpcone.c.

Referenced by DSDPAddCone(), DSDPConeDestroy(), and DSDPDestroyCones().

int DSDPConeInvertS ( DSDPCone K )

Invert the dual matrix S.

Parameters:

K

the cone

Assumes that the matrix has already been factored.

Definition at line 265 of file dsdpcone.c.

Referenced by DSDPInvertS().

int DSDPConeMonitor	(	DSDPCone	K,
		int	tag
	)

Do anything at in the cone at each iteration.

Parameters:

	K	the cone
	tag	allows for multiple types of monitors.

This routine has be used to visualize data, print some statistics, ...

Definition at line 380 of file dsdpcone.c.

Referenced by DSDPMonitorCones().

int DSDPConeMultiplyAdd	(	DSDPCone	K,
		double	mu,
		DSDPVec	vrow,
		DSDPVec	v,
		DSDPVec	vv
	)

Multiply Hessian by a vector and add the result.

Parameters:

	K	the cone
	mu	barrier parameter
	vrow	scaling for each element in the product.
	v	input vector gradient
	vv	output vector

This routine assumes that the dual matrix has already been factored and inverted. If M is the hessian, then vv += vrow .* Mv

Definition at line 119 of file dsdpcone.c.

Referenced by DSDPHessianMultiplyAdd().

int DSDPConeSetData	(	DSDPCone *	K,
		struct DSDPCone_Ops *	ops,
		void *	data
	)

Initialize the pointers to 0.

Parameters:

	K	the cone
	ops	address of a structure of function pointers.
	data	address of a structure representing a cone

Definition at line 477 of file dsdpcone.c.

Referenced by DSDPAddCone(), and DSDPConeInitialize().

int DSDPConeSetUp	(	DSDPCone	K,
		DSDPVec	y
	)

Factor the data and allocate data structures.

Parameters:

	K	the cone
	y	initial solution vector

Definition at line 22 of file dsdpcone.c.

Referenced by DSDPSetUpCones().

int DSDPConeSetUp2	(	DSDPCone	K,
		DSDPVec	yy0,
		DSDPSchurMat	M
	)

Factor the data and allocate data structures.

Parameters:

	K	the cone
	yy0	initial solution vector
	M	Schur matrix

Definition at line 43 of file dsdpcone.c.

Referenced by DSDPSetUpCones2().

int DSDPConeSetXMaker	(	DSDPCone	K,
		double	mu,
		DSDPVec	y,
		DSDPVec	dy
	)

Pass information needed to construct X.

Parameters:

	K	the cone
	mu	barrier parameter
	y	solution
	dy	step direction

Definition at line 191 of file dsdpcone.c.

Referenced by DSDPPassXVectors().

int DSDPConeSparsityInSchurMat	(	DSDPCone	K,
		int	row,
		int	rnnz[],
		int	m
	)

Identify sparsity pattern in a row of the Hessian term.

Parameters:

	K	the cone
	row	between 1 and m
	rnnz	mark elements nonzero for nonzeros in Hessian of barrier.
	m	number of y variables, length of array, and size of M matrix

See also:: DSDPSparsityInSchurMat()
DSDPSchurSparsity()

Definition at line 338 of file dsdpcone.c.

Referenced by DSDPSchurSparsity().

int DSDPConeView ( DSDPCone K )

View contents of the cone.

Parameters:

K

the cone

Definition at line 358 of file dsdpcone.c.

Referenced by DSDPViewCones().

int DSDPGetConeName	(	DSDPCone	K,
		char *	cname,
		int	maxlength
	)

Get name of the cone.

Parameters:

	K	the cone
	cname	string to copy the string
	maxlength	maximum length of the string.

Definition at line 427 of file dsdpcone.c.

Referenced by DSDPSetCone().


Typedefs
typedef struct DSDPCone_C	DSDPCone
	This object holds the data of a SDP, LP, or other cone. Its structure is opaque to the DSDP Solver, but it must implement the interface below and provide a structure of function pointers.
Functions
int	DSDPConeANorm2 (DSDPCone, DSDPVec)
	Add square of 2-norm of data correponding to each variable y.
int	DSDPConeComputeHessian (DSDPCone, double, DSDPSchurMat, DSDPVec, DSDPVec)
	Compute Hessian and gradient of barrier function.
int	DSDPConeComputeLogSDeterminant (DSDPCone, double , double )
	Evaluate logrithmic barrier function.
int	DSDPConeComputeMaxStepLength (DSDPCone, DSDPVec, DSDPDualFactorMatrix, double *)
	Determine distance to the edge of the cone.
int	DSDPConeComputeRHS (DSDPCone, double, DSDPVec, DSDPVec, DSDPVec)
	Compute gradient of barrier function.
int	DSDPConeComputeS (DSDPCone, DSDPVec, DSDPDualFactorMatrix, DSDPTruth *)
	Given y, compute S and determine whether its in the cone.
int	DSDPConeComputeX (DSDPCone, double, DSDPVec, DSDPVec, DSDPVec, double *)
	Given y,dy, and mu, construct X and add its inner product with the data and S.
int	DSDPConeDestroy (DSDPCone *)
	Free the internal memory of the cone.
int	DSDPConeGetDimension (DSDPCone, double *)
	Provide the dimension of the cone.
int	DSDPConeInitialize (DSDPCone *)
	Initialize the pointers to 0.
int	DSDPConeInvertS (DSDPCone)
	Invert the dual matrix S.
int	DSDPConeMonitor (DSDPCone, int)
	Do anything at in the cone at each iteration.
int	DSDPConeMultiplyAdd (DSDPCone, double, DSDPVec, DSDPVec, DSDPVec)
	Multiply Hessian by a vector and add the result.
int	DSDPConeSetData (DSDPCone , struct DSDPCone_Ops , void *)
	Initialize the pointers to 0.
int	DSDPConeSetUp (DSDPCone, DSDPVec)
	Factor the data and allocate data structures.
int	DSDPConeSetUp2 (DSDPCone, DSDPVec, DSDPSchurMat)
	Factor the data and allocate data structures.
int	DSDPConeSetXMaker (DSDPCone, double, DSDPVec, DSDPVec)
	Pass information needed to construct X.
int	DSDPConeSparsityInSchurMat (DSDPCone, int, int[], int)
	Identify sparsity pattern in a row of the Hessian term.
int	DSDPConeView (DSDPCone)
	View contents of the cone.
int	DSDPGetConeName (DSDPCone, char *, int)
	Get name of the cone.