Class LineSearch

This line search algorithm is algorithm 3.5/3.6 from Nocedal and Wright (Second edition, 2006). It provides a step length that satisfies the strong Wolfe conditions.

Inheritance

Object

LineSearch

Inherited Members

Object.Equals(Object)

Object.Equals(Object, Object)

Object.GetHashCode()

Object.GetType()

Object.MemberwiseClone()

Object.ReferenceEquals(Object, Object)

Object.ToString()

Namespace: Microsoft.ML.Probabilistic.Math

Assembly: Microsoft.ML.Probabilistic.dll

Syntax

public class LineSearch

Properties

ExtrapolationMultiplier

Multiplier for the extrapolation

Declaration

public double ExtrapolationMultiplier { get; set; }

Property Value

Type	Description
Double

InterpolationTolerance

When performing interpolation this is the minimum fraction of the range by which the interpolant must be from one of the two end points

Declaration

public double InterpolationTolerance { get; set; }

Property Value

Type	Description
Double

MaximumZoomIterations

Maximum number of iterations for the zoom method

Declaration

public int MaximumZoomIterations { get; set; }

Property Value

Type	Description
Int32

Phi

Line search evaluation function

Declaration

public LineSearchEval Phi { get; set; }

Property Value

Type	Description
LineSearchEval

Methods

Run(Double, Double, Double, Double)

Main execution method for line search

Declaration

public double Run(double initialStep, double maxStep, double phi0, double dPhi0)

Parameters

Type	Name	Description
Double	initialStep
Double	maxStep
Double	phi0
Double	dPhi0

Returns

Type	Description
Double	Step satisfying Wolfe conditions

SetWolfeConstants(Double, Double)

Set Wolfe conditions

Declaration

public void SetWolfeConstants(double c1, double c2)

Parameters

Type	Name	Description
Double	c1	Sufficient decrease condition
Double	c2	Curvature condition