Package org.ddogleg.optimization.trustregion

Class TrustRegionLeastSqBase_F64<S extends DMatrix,HM extends HessianMath>

java.lang.Object

org.ddogleg.optimization.GaussNewtonBase_F64<ConfigTrustRegion,HM>

org.ddogleg.optimization.trustregion.TrustRegionBase_F64<S,HM>

org.ddogleg.optimization.trustregion.TrustRegionLeastSqBase_F64<S,HM>

All Implemented Interfaces:

VerbosePrint

Direct Known Subclasses:

UnconLeastSqTrustRegion_F64, UnconLeastSqTrustRegionSchur_F64

public abstract class TrustRegionLeastSqBase_F64<S extends DMatrix,HM extends HessianMath>
extends TrustRegionBase_F64<S,HM>

Base class for all least squares trust region implementations.

Nested Class Summary

Nested classes/interfaces inherited from class org.ddogleg.optimization.trustregion.TrustRegionBase_F64

TrustRegionBase_F64.ParameterUpdate<S extends DMatrix>

Nested classes/interfaces inherited from class org.ddogleg.optimization.GaussNewtonBase_F64

GaussNewtonBase_F64.Mode

Field Summary

Fields

Modifier and Type	Field	Description
protected FunctionNtoM	functionResiduals
protected LossFunction	lossFunc	Given the residuals it computes the "Loss" or cost
protected @Nullable LossFunctionGradient	lossFuncGradient	Gradient of the loss function.
protected DMatrixRMaj	residuals
protected DMatrixRMaj	storageLossGradient

Fields inherited from class org.ddogleg.optimization.trustregion.TrustRegionBase_F64

gradientNorm, parameterUpdate, tmp_p

Fields inherited from class org.ddogleg.optimization.GaussNewtonBase_F64

config, ftest_val, fx, gradient, gtest_val, hessian, hessianScaling, mode, p, postUpdateAdjuster, sameStateAsCost, totalFullSteps, totalSelectSteps, verbose, verboseLevel, x, x_next

Constructor Summary

Constructors

Modifier	Constructor	Description
protected	TrustRegionLeastSqBase_F64(TrustRegionBase_F64.ParameterUpdate<S> parameterUpdate, HM hessian)

Method Summary

Modifier and Type	Method	Description
protected boolean	acceptNewState(boolean converged, double fx_candidate)
protected double	computeCostAtInitialization()	This function was created for least squares which requires special set up due to the Loss function
protected double	cost(DMatrixRMaj x)	Computes the function's value at x
void	setLoss(LossFunction loss, LossFunctionGradient lossGradient)	Specifies the loss function.

Methods inherited from class org.ddogleg.optimization.trustregion.TrustRegionBase_F64

checkConvergenceFTest, computeStep, configure, considerCandidate, getConfig, initialize, setVerbose, solveCauchyStepLength, updateDerivates

Methods inherited from class org.ddogleg.optimization.GaussNewtonBase_F64

applyHessianScaling, checkConvergenceGTest, computeHessianScaling, computeHessianScaling, computePredictedReduction, functionGradientHessian, initialize, iterate, mode, setVerbose, undoHessianScalingOnParameters

Methods inherited from class java.lang.Object

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

Field Details

functionResiduals

protected FunctionNtoM functionResiduals

residuals

protected DMatrixRMaj residuals

lossFunc

protected LossFunction lossFunc

Given the residuals it computes the "Loss" or cost

lossFuncGradient

@Nullable
protected @Nullable LossFunctionGradient lossFuncGradient

Gradient of the loss function. If null then squared error is assumed and this step can be skipped.

storageLossGradient

protected DMatrixRMaj storageLossGradient

Constructor Details

TrustRegionLeastSqBase_F64

protected TrustRegionLeastSqBase_F64(TrustRegionBase_F64.ParameterUpdate<S> parameterUpdate,
 HM hessian)

Method Details

setLoss

public void setLoss(LossFunction loss,
 LossFunctionGradient lossGradient)

Specifies the loss function.

computeCostAtInitialization

protected double computeCostAtInitialization()

Description copied from class: TrustRegionBase_F64

This function was created for least squares which requires special set up due to the Loss function

Overrides:

computeCostAtInitialization in class TrustRegionBase_F64<S extends DMatrix,HM extends HessianMath>

acceptNewState

protected boolean acceptNewState(boolean converged,
 double fx_candidate)

Overrides:

acceptNewState in class TrustRegionBase_F64<S extends DMatrix,HM extends HessianMath>

cost

protected double cost(DMatrixRMaj x)

Description copied from class: TrustRegionBase_F64

Computes the function's value at x

Specified by:

cost in class TrustRegionBase_F64<S extends DMatrix,HM extends HessianMath>

Parameters:

x - parameters

Returns:

function value