org.apache.flink.ml.optimization

Class GradientDescentL1

java.lang.Object

org.apache.flink.ml.optimization.Solver

org.apache.flink.ml.optimization.IterativeSolver

org.apache.flink.ml.optimization.GradientDescent

org.apache.flink.ml.optimization.GradientDescentL1

All Implemented Interfaces:

Serializable, WithParameters

public class GradientDescentL1
extends GradientDescent

Implementation of a SGD solver with L1 regularization.

The regularization function is ||w||_1 with w being the weight vector.

See Also:

Serialized Form

Nested Class Summary

Nested classes/interfaces inherited from class org.apache.flink.ml.optimization.IterativeSolver

IterativeSolver.ConvergenceThreshold$, IterativeSolver.Iterations$, IterativeSolver.LearningRate$, IterativeSolver.LearningRateMethodValue$

Nested classes/interfaces inherited from class org.apache.flink.ml.optimization.Solver

Solver.LossFunction$, Solver.RegularizationConstant$

Constructor Summary

Constructors

Constructor and Description
GradientDescentL1()

Method Summary

Modifier and Type	Method and Description
static GradientDescentL1	apply()
Vector	takeStep(Vector weightVector, Vector gradient, double regularizationConstant, double learningRate) Calculates the new weights based on the gradient.

Methods inherited from class org.apache.flink.ml.optimization.GradientDescent

optimize, optimizeWithConvergenceCriterion, optimizeWithoutConvergenceCriterion

Methods inherited from class org.apache.flink.ml.optimization.IterativeSolver

MAX_DLOSS, setConvergenceThreshold, setIterations, setLearningRateMethod, setStepsize

Methods inherited from class org.apache.flink.ml.optimization.Solver

createInitialWeightsDS, createInitialWeightVector, setLossFunction, setRegularizationConstant

Methods inherited from class java.lang.Object

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

Methods inherited from interface org.apache.flink.ml.common.WithParameters

parameters

Constructor Detail

GradientDescentL1

public GradientDescentL1()

Method Detail

apply

public static GradientDescentL1 apply()

takeStep

public Vector takeStep(Vector weightVector,
                       Vector gradient,
                       double regularizationConstant,
                       double learningRate)

Calculates the new weights based on the gradient.

Specified by:

takeStep in class GradientDescent

Parameters:

weightVector -

gradient -

regularizationConstant -

learningRate -

Returns: