Package ch.idsia.crema.factor.bayesian

Interface BayesianFactor

All Superinterfaces:

FilterableFactor<BayesianFactor>, GenericFactor, OperableFactor<BayesianFactor>, SeparatelySpecified<BayesianFactor>

All Known Implementing Classes:

BayesianAbstractFactor, BayesianAndFactor, BayesianDefaultFactor, BayesianDeterministicFactor, BayesianFunctionFactor, BayesianFunctionLogFactor, BayesianLogFactor, BayesianLogicFactor, BayesianNoisyOrFactor, BayesianNotFactor, BayesianOrFactor

public interface BayesianFactor
extends OperableFactor<BayesianFactor>, SeparatelySpecified<BayesianFactor>

Author: Claudio "Dna" Bonesana Project: crema Date: 17.03.2021 13:18

Method Summary

Modifier and Type	Method	Description
BayesianFactor	combine(BayesianFactor other)	Combine this factor with the provided one and return the result as a new factor.
BayesianFactor	copy()
BayesianFactor	divide(BayesianFactor factor)
BayesianFactor	filter(int variable, int state)	Filter the factor by selecting only the values where the specified variable is in the specified state.
double[]	getData()
Strides	getDataDomain()	The domain of the rest of the domain.
Strides	getDomain()
double	getLogValue(int... states)
double	getLogValueAt(int index)
Strides	getSeparatingDomain()	The domain of the separated part.
double	getValue(int... states)
double	getValueAt(int index)
double	logProb(gnu.trove.map.TIntIntMap[] data, int leftVar)
BayesianFactor	marginalize(int variable)	Sum out a variable from the factor.
ObservationBuilder	sample()

Methods inherited from interface ch.idsia.crema.factor.FilterableFactor

filter

Methods inherited from interface ch.idsia.crema.factor.GenericFactor

isLog

Methods inherited from interface ch.idsia.crema.factor.OperableFactor

combine, combine, marginalize, normalize

Method Detail

getDomain

Strides getDomain()

Specified by:

getDomain in interface GenericFactor

Returns:

The domain of the factor. This includes variables, sizes and strides.

getDataDomain

Strides getDataDomain()

Description copied from interface: SeparatelySpecified

The domain of the rest of the domain. Together with SeparatelySpecified.getSeparatingDomain() it will form the complete domain of the object.

Specified by:

getDataDomain in interface SeparatelySpecified<BayesianFactor>

Returns:

getSeparatingDomain

Strides getSeparatingDomain()

Description copied from interface: SeparatelySpecified

The domain of the separated part. This is usually the conditioning of the Factor. This side must not overlap with the data domain returned by SeparatelySpecified.getDataDomain().

Specified by:

getSeparatingDomain in interface SeparatelySpecified<BayesianFactor>

Returns:

getValue

double getValue(int... states)

Parameters:

states - specified in the order of the factor's domain

Returns:

the value associated to the given states

getValueAt

double getValueAt(int index)

Parameters:

index - offset based on the factor's domain

Returns:

the value associated to the given states

getLogValue

double getLogValue(int... states)

Parameters:

states - specified in the order of the factor's domain

Returns:

the value associated to the given states in log format

getLogValueAt

double getLogValueAt(int index)

Parameters:

index - offset based on the factor's domain

Returns:

the value associated to the given states in log format

getData

double[] getData()

copy

BayesianFactor copy()

Specified by:

copy in interface GenericFactor

Returns:

a copy of the factor

filter

BayesianFactor filter(int variable,
                      int state)

Description copied from interface: OperableFactor

Filter the factor by selecting only the values where the specified variable is in the specified state.

Should return this if the variable is not part of the domain of the factor.

Specified by:

filter in interface FilterableFactor<BayesianFactor>

Specified by:

filter in interface OperableFactor<BayesianFactor>

Parameters:

variable - variable to filter

state - state of the variable to filter

Returns:

a new factor where the given variable in the given state has been filtered out

combine

BayesianFactor combine(BayesianFactor other)

Description copied from interface: OperableFactor

Combine this factor with the provided one and return the result as a new factor.

Specified by:

combine in interface OperableFactor<BayesianFactor>

Parameters:

other - other factor to combine with this one

Returns:

a new factor combination of this with the given one

marginalize

BayesianFactor marginalize(int variable)

Description copied from interface: OperableFactor

Sum out a variable from the factor.

Specified by:

marginalize in interface OperableFactor<BayesianFactor>

Parameters:

variable - variable to be marginalize out from this factor

Returns:

a new factor without the given variable

divide

BayesianFactor divide(BayesianFactor factor)

Specified by:

divide in interface OperableFactor<BayesianFactor>

Parameters:

factor - given factor to divide by

Returns:

a new factor resulting in the division of this factor by the given one

sample

ObservationBuilder sample()

logProb

double logProb(gnu.trove.map.TIntIntMap[] data,
               int leftVar)