Detailed Description

Interface for an error function returning one or multiple values.

Public Member Functions
virtual void	evaluateError (std::vector< double > &out)=0
	Evaluate the error function.

virtual double	penalize (double cost, const std::vector< double > &error)
	Optionally override this method to penalize a cost function based on the error function result.

Static penalty tool methods that are useful elsewhere
static double	errorToMultPenalty (double value, double range)
	There is a margin of half the range within which the returned penalty value fades from 1 to 0 with a polynomial function, and extents below zero after 1.5x range, and is continuous in the derivative.

static double	errorsToMultPenalty (const std::vector< double > &error, const std::vector< double > &range)
	Return a number of error values to a combined penalty function, while properly housekeeping values < 0.

Member Function Documentation

◆ evaluateError()

virtual void evaluateError ( std::vector< double > & out )

pure virtual

Evaluate the error function.

An implementation of this method shall resize the out array and store some values in there, based on the most recent evaluation of a cost function associated with it.

Implemented in PointCorr, SweepGlobalInitAlgorithm, and SweepPointCorrAlgorithm.

◆ penalize()

virtual double penalize	(	double	cost,
		const std::vector< double > &	error )

inlinevirtual

Optionally override this method to penalize a cost function based on the error function result.

Reimplemented in SweepGlobalInitAlgorithm, and SweepPointCorrAlgorithm.

◆ errorToMultPenalty()

static double errorToMultPenalty	(	double	value,
		double	range )

static

There is a margin of half the range within which the returned penalty value fades from 1 to 0 with a polynomial function, and extents below zero after 1.5x range, and is continuous in the derivative.

Convert error value to a multiplicative penalty value, returning a value < 1 if the range is exceeded

◆ errorsToMultPenalty()

static double errorsToMultPenalty	(	const std::vector< double > &	error,
		const std::vector< double > &	range )

static

Return a number of error values to a combined penalty function, while properly housekeeping values < 0.

If one or more of the error values would result in a negative penalty, those values are added up and the remaining positive ones ignored; all positive ones are multiplied together.

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

ImFusion/Base/ErrorFunction.h