hex.glm

Class GLM.ProximalGradientSolver

java.lang.Object

hex.glm.GLM.ProximalGradientSolver

All Implemented Interfaces:

ADMM.ProximalSolver, OptimizationUtils.GradientSolver

Enclosing class:

GLM

public static class GLM.ProximalGradientSolver
extends java.lang.Object
implements OptimizationUtils.GradientSolver, ADMM.ProximalSolver

Simple wrapper around ginfo computation, adding proximal penalty

Constructor Summary

Constructors

Constructor and Description
GLM.ProximalGradientSolver(OptimizationUtils.GradientSolver s, double[] betaStart, double[] rho, double objEps, double gradEps, OptimizationUtils.GradientInfo ginfo, L_BFGS.ProgressMonitor pm)

Method Summary

Methods

Modifier and Type	Method and Description
GLM.ProximalGradientInfo	getGradient(double[] beta) Evaluate ginfo at solution beta.
OptimizationUtils.GradientInfo	getObjective(double[] beta)
OptimizationUtils.GradientInfo	gradient(double[] beta)
boolean	hasGradient()
int	iter()
static double	proximal_gradient(double[] grad, double obj, double[] beta, double[] beta_given, double[] rho)
double[]	rho()
boolean	solve(double[] beta_given, double[] beta)

Methods inherited from class java.lang.Object

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

Constructor Detail

GLM.ProximalGradientSolver

public GLM.ProximalGradientSolver(OptimizationUtils.GradientSolver s,
                          double[] betaStart,
                          double[] rho,
                          double objEps,
                          double gradEps,
                          OptimizationUtils.GradientInfo ginfo,
                          L_BFGS.ProgressMonitor pm)

Method Detail

proximal_gradient

public static double proximal_gradient(double[] grad,
                       double obj,
                       double[] beta,
                       double[] beta_given,
                       double[] rho)

getGradient

public GLM.ProximalGradientInfo getGradient(double[] beta)

Description copied from interface: OptimizationUtils.GradientSolver

Evaluate ginfo at solution beta.

Specified by:

getGradient in interface OptimizationUtils.GradientSolver

Returns:

getObjective

public OptimizationUtils.GradientInfo getObjective(double[] beta)

Specified by:

getObjective in interface OptimizationUtils.GradientSolver

rho

public double[] rho()

Specified by:

rho in interface ADMM.ProximalSolver

solve

public boolean solve(double[] beta_given,
            double[] beta)

Specified by:

solve in interface ADMM.ProximalSolver

hasGradient

public boolean hasGradient()

Specified by:

hasGradient in interface ADMM.ProximalSolver

gradient

public OptimizationUtils.GradientInfo gradient(double[] beta)

Specified by:

gradient in interface ADMM.ProximalSolver

iter

public int iter()

Specified by:

iter in interface ADMM.ProximalSolver