gov.sandia.cognition.learning.function.cost

Class NegativeLogLikelihood<DataType>

java.lang.Object

gov.sandia.cognition.util.AbstractCloneableSerializable

gov.sandia.cognition.learning.function.cost.AbstractCostFunction<ComputableDistribution<DataType>,java.util.Collection<? extends DataType>>

gov.sandia.cognition.learning.function.cost.NegativeLogLikelihood<DataType>

Type Parameters:

DataType - Type of data generated by the Distribution

All Implemented Interfaces:

Evaluator<ComputableDistribution<DataType>,java.lang.Double>, CostFunction<ComputableDistribution<DataType>,java.util.Collection<? extends DataType>>, CloneableSerializable, java.io.Serializable, java.lang.Cloneable

Direct Known Subclasses:

ParallelNegativeLogLikelihood

public class NegativeLogLikelihood<DataType>
extends AbstractCostFunction<ComputableDistribution<DataType>,java.util.Collection<? extends DataType>>

CostFunction for computing the maximum likelihood (because we are minimizing the negative of the log likelihood)

See Also:

Serialized Form

Field Summary

Fields inherited from class gov.sandia.cognition.learning.function.cost.AbstractCostFunction

costParameters

Constructor Summary

Constructors

Constructor and Description
NegativeLogLikelihood() Default constructor
NegativeLogLikelihood(java.util.Collection<? extends DataType> costParameters) Creates a new instance of NegativeLogLikelihood

Method Summary

Modifier and Type	Method and Description
java.lang.Double	evaluate(ComputableDistribution<DataType> target) Computes the cost of the given target.
static <DataType> double	evaluate(ProbabilityFunction<DataType> f, java.util.Collection<? extends DataType> data) Evaluates the negative log-likelihood of the given collection of data according to the given probability function.

Methods inherited from class gov.sandia.cognition.learning.function.cost.AbstractCostFunction

clone, getCostParameters, setCostParameters

Methods inherited from class java.lang.Object

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

Constructor Detail

NegativeLogLikelihood

public NegativeLogLikelihood()

Default constructor

NegativeLogLikelihood

public NegativeLogLikelihood(java.util.Collection<? extends DataType> costParameters)

Creates a new instance of NegativeLogLikelihood

Parameters:

costParameters - Data generated by the target distribution

Method Detail

evaluate

public java.lang.Double evaluate(ComputableDistribution<DataType> target)

Description copied from interface: CostFunction

Computes the cost of the given target.

Parameters:

target - The object to evaluate.

Returns:

The cost of the given object.

evaluate

public static <DataType> double evaluate(ProbabilityFunction<DataType> f,
                                         java.util.Collection<? extends DataType> data)

Evaluates the negative log-likelihood of the given collection of data according to the given probability function.

Type Parameters:

DataType - The type of data generated by the distribution.

Parameters:

f - The function to compute the log-likelihood.

data - The data to compute the log-likelihood of.

Returns:

The total negative log-likelihood of the data according to the function.