Package org.ddogleg.optimization.math

Interface HessianSchurComplement<S extends DMatrix>

All Superinterfaces:

HessianMath

All Known Implementing Classes:

HessianSchurComplement_Base, HessianSchurComplement_DDRM, HessianSchurComplement_DSCC

public interface HessianSchurComplement<S extends DMatrix>
extends HessianMath

Given the already computed Jacobian (broken up into a left and right side) compute the decomposed approximate Hessian matrix, i.e. [A B, B D]

     A=L'*L
     B=L'*R
     D=R*R
 

Where L and R are the left and right hand side of the Jacobian, respectively

Method Summary

Modifier and Type	Method	Description
void	computeGradient(S jacLeft, S jacRight, DMatrixRMaj residuals, DMatrixRMaj gradient)	Computes the gradient given the Jacobian and the residuals.
void	computeHessian(S jacLeft, S jacRight)	Given the left and right hand side of the Jacobian compute the Hessian.
S	createMatrix()	Creates a matrix of the same type that this interface can process

Methods inherited from interface org.ddogleg.optimization.math.HessianMath

divideRowsCols, extractDiagonals, init, initializeSolver, innerVectorHessian, setDiagonals, solve

Method Details

computeHessian

void computeHessian(S jacLeft,
 S jacRight)

Given the left and right hand side of the Jacobian compute the Hessian.

Parameters:

jacLeft - (input) Jacobian left side

jacRight - (input) Jacobian right side

computeGradient

void computeGradient(S jacLeft,
 S jacRight,
 DMatrixRMaj residuals,
 DMatrixRMaj gradient)

Computes the gradient given the Jacobian and the residuals.

Parameters:

jacLeft - (input) Jacobian left side

jacRight - (input) Jacobian right side

residuals - (Input) residuals

gradient - (Output) gradient

createMatrix

S createMatrix()

Creates a matrix of the same type that this interface can process

Returns:

matrx