vec_math
Class HarmonicFit.HarmonicModel

java.lang.Object
  vec_math.AbstractDataModel
      vec_math.AbstractGradientModel
          vec_math.HarmonicFit.HarmonicModel

All Implemented Interfaces:

DataModel, GradientModel

Enclosing class:

HarmonicFit

private static class HarmonicFit.HarmonicModel
extends AbstractGradientModel

Our data model.

Nested Class Summary

Nested classes/interfaces inherited from class vec_math.AbstractDataModel

AbstractDataModel.Construct, AbstractDataModel.LocalM

Field Summary

(package private) double	external Square of external error.
(package private) boolean	fitting If true, we return the almost-linear model for LM-fitting.

Fields inherited from class vec_math.AbstractDataModel

sigma, x, y

Constructor Summary

protected

HarmonicFit.HarmonicModel(VectorG[] times, double[] y, double[] err, double ext2)

Method Summary

double	evaluateModel(VectorG tup4, VectorG x1) The model parameters are per index 0: frequency f = 1/P; ω=2πf 1: cosine amplitude 2: sine amplitude 3: constant offset
Matrix	getGradientMatrix(VectorG tup4) The gradient matrix of the model holds per column index 0: d/df frequency f = 1/P; ω=2πf.
double[]	getMeasurementErrors() If we had an external error, we subtract it again.
int	getParameterCount() The number of model parameters should be less than the number of data points in the model, otherwise you fit a underdetermined problem.
void	setFitting(boolean fit) Toggle the fitting flag.

Methods inherited from class vec_math.AbstractGradientModel

check, check, getNegativeChi2HalfGradient, getNegativeChi2HalfGradient, getPseudoHessian, getPseudoHessian, LevenbergMarquardtSolver

Methods inherited from class vec_math.AbstractDataModel

getChiSquareModel, getChiSquareModel, getLorentzianModel, getMeasureCount, getMeasures, getModel, getModel, getResiduals, getResiduals, getRms, getRobustModel, getTimes, getTotalErrors, LorentzianSolver, RobustSolver, SimplexSolver

Methods inherited from class java.lang.Object

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

Methods inherited from interface vec_math.DataModel

getMeasureCount, getMeasures, getModel, getResiduals, getTimes, getTotalErrors

Field Detail

external

double external

Square of external error.

fitting

boolean fitting

If true, we return the almost-linear model for LM-fitting.

Constructor Detail

HarmonicFit.HarmonicModel

protected HarmonicFit.HarmonicModel(VectorG[] times,
                                    double[] y,
                                    double[] err,
                                    double ext2)

Method Detail

getParameterCount

public int getParameterCount()

Description copied from interface: DataModel

The number of model parameters should be less than the number of data points in the model, otherwise you fit a underdetermined problem.

Returns:

The number of parameters in the model. This equals the dimension of the input vector in DataModel.getModel(vec_math.VectorG).

setFitting

public void setFitting(boolean fit)

Toggle the fitting flag. During minimization with Levenberg- Marquardt or Simplex, this should be set to true. After the final solution has been converted to the human-readable form, set this to false to get correct model data.

evaluateModel

public double evaluateModel(VectorG tup4,
                            VectorG x1)

The model parameters are per index

0: frequency f = 1/P; ω=2πf 1: cosine amplitude 2: sine amplitude 3: constant offset

Parameters:

x1 - 1-d vector holding the time.

tup4 - The model parameters

getMeasurementErrors

public double[] getMeasurementErrors()

If we had an external error, we subtract it again.

Specified by:

getMeasurementErrors in interface DataModel

Overrides:

getMeasurementErrors in class AbstractDataModel

getGradientMatrix

public Matrix getGradientMatrix(VectorG tup4)

The gradient matrix of the model holds per column index

0: d/df frequency f = 1/P; ω=2πf. Evaluates to 2π*t*(-ac*sin(2πf*t)+as*cos(2πf*t)) 1: d/dac cosine amplitude, cos(2πf*t) 2: d/das sine amplitude, sin(2πf*t) 3: 1 = d/dc constant offset