Package opennlp.tools.ml.maxent.quasinewton

Class LineSearch

java.lang.Object

opennlp.tools.ml.maxent.quasinewton.LineSearch

public class LineSearch
extends Object

Performs line search to find a minimum.

See Also:

• Nocedal invalid input: '&' Wright 2006, Numerical Optimization, p. 37)

Nested Class Summary

Nested Classes

Modifier and Type	Class	Description
static class	LineSearch.LineSearchResult	Represents a LineSearch result encapsulating the relevant data at a point in time during computation.

Constructor Summary

Constructors

Constructor	Description
LineSearch()

Method Summary

Modifier and Type	Method	Description
static void	doConstrainedLineSearch(Function function, double[] direction, LineSearch.LineSearchResult lsr, double l1Cost, double initialStepSize)	Conducts a constrained line search (see section 3.2 in the paper "Scalable Training of L1-Regularized Log-Linear Models", Andrew et al. 2007)
static void	doLineSearch(Function function, double[] direction, LineSearch.LineSearchResult lsr, double initialStepSize)	Conducts a backtracking line search.

Methods inherited from class java.lang.Object

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

Constructor Details

LineSearch

public LineSearch()

Method Details

doLineSearch

public static void doLineSearch(Function function,
 double[] direction,
 LineSearch.LineSearchResult lsr,
 double initialStepSize)

Conducts a backtracking line search.

Parameters:

function - The Function to apply.

direction - The double[] representing the direction to search into.

lsr - The LineSearch.LineSearchResult to transport results in.

initialStepSize - The initial step size to apply. Must be greater than 0.

doConstrainedLineSearch

public static void doConstrainedLineSearch(Function function,
 double[] direction,
 LineSearch.LineSearchResult lsr,
 double l1Cost,
 double initialStepSize)

Conducts a constrained line search (see section 3.2 in the paper "Scalable Training of L1-Regularized Log-Linear Models", Andrew et al. 2007)

Parameters:

function - The Function to apply.

direction - The double[] representing the direction to search into.

lsr - The LineSearch.LineSearchResult to transport results in.

l1Cost - The L1-regularization costs. Must be equal or greater than 0.

initialStepSize - The initial step size to apply. Must be greater than 0.