Package pulse.search.direction

Class LMOptimiser

    • Field Detail

      • MAX_FAILED_ATTEMPTS

        public static final int MAX_FAILED_ATTEMPTS
        Up to 5 failed attempts are allowed.
        See Also:
        Constant Field Values

    • Method Detail

      • iteration

        public boolean iteration​(GeneralTask task)
                  throws SolverException
        Description copied from class: PathOptimiser

        This method sets out the basic algorithm for estimating the minimum of the target function, which is defined as the sum of squared residuals (SSR), or the deviations of the model solution (a DifferenceScheme used to solve the Problem for this task) from the empirical values (the ExperimentalData). The algorithm will go through the following steps: (1) find the direction, which points to the minimum, using the concrete direction method; (2) estimate the magnitude of the step to reach the minimum using the LinearSolver; (3) assign a new set of parameters to the SearchTask; (4) calculate the new SSR value.

        Specified by:
        iteration in class PathOptimiser
        Parameters:
        task - a SearchTask that needs to be driven to a minimum of SSR.
        Returns:
        the SSR value with the newly found parameters.
        Throws:
        SolverException
        See Also:
        direction(Path), LinearOptimiser

      • jacobian

        public RectangularMatrix jacobian​(GeneralTask task)
                           throws SolverException

        Calculates the Jacobian of the model function given as a discrete set of time-signal values. The elements of the Jacobian are calculated using central differences from two residual vectors evaluated by shifting the search vector slightly to the right or left of each search parameter.

        This is also equivalent to calculating the difference of the model values when performing the shift, when taking the model values at the time points of the reference dataset. Because of a different discretisation of the model, it is easier to substitute these with the residuals, which had already been interpolated at the reference time values.

        Parameters:
        task - the task being optimised
        Returns:
        the jacobian matrix
        Throws:
        SolverException

      • listedKeywords

        public Set<NumericPropertyKeyword> listedKeywords()
        Description copied from class: GradientBasedOptimiser

        The types of the listed parameters for this class include: GRADIENT_RESOLUTION, ERROR_TOLERANCE, ITERATION_LIMIT. Also, all the flags in this class are treated as separate listed parameters.

        Overrides:
        listedKeywords in class GradientBasedOptimiser
        Returns:
        a set of NumericPropertyKeyword instances, which have been explicitly marked as a listed parameter for this PropertyHolder.
        See Also:
        NumericPropertyKeyword

      • getInstance

        public static LMOptimiser getInstance()
        This class uses a singleton pattern, meaning there is only instance of this class.
        Returns:
        the single (static) instance of this class

      • compatibleWith

        public boolean compatibleWith​(OptimiserStatistic os)
        The Levenberg-Marquardt optimiser will only accept ordinary least-squares as its objective function. Therefore, os should be an instance of SumOfSquares.
        Overrides:
        compatibleWith in class PathOptimiser
        Parameters:
        os - a selected optimiser metric
        Returns:
        true if returns SumOfSquares.class, false otherwise

      • setDampingRatio

        public void setDampingRatio​(NumericProperty dampingRatio)