SPSA¶

public class SPSA¶

Implements SPSA optimisation algorithm (Simultaneous Perturbation Stochastic Approximation). This version optimises the cells of the OD matrix. http://www.jhuapl.edu/SPSA/

Author:	Milan Lovric

Fields¶

THETA_MAX¶

public static final double THETA_MAX¶

THETA_MIN¶

public static final double THETA_MIN¶

Constructors¶

SPSA¶

public SPSA()¶

Methods¶

getLossFunctionEvaluations¶

public List<Double> getLossFunctionEvaluations()¶

Returns:	Loss function evaluations for all iterations.

getThetaEstimate¶

public RealODMatrix getThetaEstimate()¶

Getter function for the optimisation result (OD matrix).

Returns:	Estimated OD matrix.

initialise¶

public void initialise(RoadNetworkAssignment rna, Properties props, RealODMatrix initialTheta, double a, double A, double c, double alpha, double gamma)¶

Initialise the SPSA algorithm with starting values.

Parameters:	rna – Road network assignment. props – Parameters from the config file. initialTheta – Initial OD matrix. a – SPSA parameter. A – SPSA parameter. c – SPSA parameter. alpha – SPSA parameter. gamma – SPSA parameter.

lossFunction¶

public double lossFunction()¶

Calculate the loss function of the latest theta estimate (OD matrix).

Returns:	RMSN for the difference between volumes and traffic counts.

runSPSA¶

public void runSPSA(int maxIterations)¶

Run the algorithm.

Parameters:	maxIterations – Maximum number of iterations.