Class LeastSquaresConverter

  • All Implemented Interfaces:
    MultivariateFunction

    public class LeastSquaresConverter
    extends java.lang.Object
    implements MultivariateFunction
    This class converts vectorial objective functions to scalar objective functions when the goal is to minimize them.
    This class is mostly used when the vectorial objective function represents a theoretical result computed from a point set applied to a model and the models point must be adjusted to fit the theoretical result to some reference observations. The observations may be obtained for example from physical measurements whether the model is built from theoretical considerations.
    This class computes a possibly weighted squared sum of the residuals, which is a scalar value. The residuals are the difference between the theoretical model (i.e. the output of the vectorial objective function) and the observations. The class implements the MultivariateFunction interface and can therefore be minimized by any optimizer supporting scalar objectives functions.This is one way to perform a least square estimation. There are other ways to do this without using this converter, as some optimization algorithms directly support vectorial objective functions.
    This class support combination of residuals with or without weights and correlations.
    Since:
    2.0
    See Also:
    MultivariateFunction, MultivariateVectorFunction
    • Field Summary

      Fields 
      Modifier and Type Field Description
      private MultivariateVectorFunction function
      Underlying vectorial function.
      private double[] observations
      Observations to be compared to objective function to compute residuals.
      private RealMatrix scale
      Optional scaling matrix (weight and correlations) for the residuals.
      private double[] weights
      Optional weights for the residuals.
    • Method Summary

      All Methods Instance Methods Concrete Methods 
      Modifier and Type Method Description
      double value​(double[] point)
      Compute the value for the function at the given point.
      • Methods inherited from class java.lang.Object

        clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait
    • Field Detail

      • observations

        private final double[] observations
        Observations to be compared to objective function to compute residuals.
      • weights

        private final double[] weights
        Optional weights for the residuals.
      • scale

        private final RealMatrix scale
        Optional scaling matrix (weight and correlations) for the residuals.
    • Constructor Detail

      • LeastSquaresConverter

        public LeastSquaresConverter​(MultivariateVectorFunction function,
                                     double[] observations)
        Builds a simple converter for uncorrelated residuals with identical weights.
        Parameters:
        function - vectorial residuals function to wrap
        observations - observations to be compared to objective function to compute residuals
      • LeastSquaresConverter

        public LeastSquaresConverter​(MultivariateVectorFunction function,
                                     double[] observations,
                                     double[] weights)
        Builds a simple converter for uncorrelated residuals with the specified weights.

        The scalar objective function value is computed as:

         objective = ∑weighti(observationi-objectivei)2
         

        Weights can be used for example to combine residuals with different standard deviations. As an example, consider a residuals array in which even elements are angular measurements in degrees with a 0.01° standard deviation and odd elements are distance measurements in meters with a 15m standard deviation. In this case, the weights array should be initialized with value 1.0/(0.012) in the even elements and 1.0/(15.02) in the odd elements (i.e. reciprocals of variances).

        The array computed by the objective function, the observations array and the weights array must have consistent sizes or a DimensionMismatchException will be triggered while computing the scalar objective.

        Parameters:
        function - vectorial residuals function to wrap
        observations - observations to be compared to objective function to compute residuals
        weights - weights to apply to the residuals
        Throws:
        DimensionMismatchException - if the observations vector and the weights vector dimensions do not match (objective function dimension is checked only when the value(double[]) method is called)
      • LeastSquaresConverter

        public LeastSquaresConverter​(MultivariateVectorFunction function,
                                     double[] observations,
                                     RealMatrix scale)
        Builds a simple converter for correlated residuals with the specified weights.

        The scalar objective function value is computed as:

         objective = yTy with y = scale×(observation-objective)
         

        The array computed by the objective function, the observations array and the the scaling matrix must have consistent sizes or a DimensionMismatchException will be triggered while computing the scalar objective.

        Parameters:
        function - vectorial residuals function to wrap
        observations - observations to be compared to objective function to compute residuals
        scale - scaling matrix
        Throws:
        DimensionMismatchException - if the observations vector and the scale matrix dimensions do not match (objective function dimension is checked only when the value(double[]) method is called)
    • Method Detail

      • value

        public double value​(double[] point)
        Compute the value for the function at the given point.
        Specified by:
        value in interface MultivariateFunction
        Parameters:
        point - Point at which the function must be evaluated.
        Returns:
        the function value for the given point.