com.bayesia.api.learning.settings

Interface LearningSettings

All Superinterfaces:

java.lang.Cloneable

All Known Implementing Classes:

AbstractLearningSettings, AugmentedMarkovBlanketSettings, AugmentedNaiveSettings, EQSettings, GreedySettings, MarkovBlanketSettings, MaximumSpanningTreeSettings, NaiveSettings, ParameterEstimationSettings, SopLEQSettings, SupervisedLearningSettings, TabooOrderSettings, TabooSettings, TreeAugmentedNaiveSettings

public interface LearningSettings
extends java.lang.Cloneable

This interface is used to describe the settings for a learning algorithm.

Each learning algorithm has a specific settings class that indicates which algorithm to use to the following method: APILearning.performLearning(LearningSettings) Each class of settings has a method createDefault that create an instance of the corresponding learning settings with the correct default parameters. Once the instance is created, it is possible to modify the values of these suttings up to specific methods.

Version:

4.2

Author:

Bayesia S.A.S.

Method Summary

Modifier and Type	Method and Description
LearningSettings	copy() Make a deep copy of the learning settings.
APIUtils.LEARNING_ALGORITHM	getLearningAlgorithm() Return a LEARNING_ALGORITHM indicating the learning algorithm to use
double	getSmoothedProbabilityEstimation() Returns the weight of the non informative prior, i.e. the virtual case(s) spread uniformly across the joint probability distribution.
double	getStructuralCoefficient() Returns the weight associated to the structural part used in the Minimum Description Length score.
java.lang.String	getTargetNodeName() Return the name of the target node if the learning algorithm is supervised.
boolean	isSupervised() Indicate if the current learning algorithm is supervised (i.e. it needs a target) or not.
boolean	isValid() Test if the current settings are valid.
boolean	isVerbose() Tests if the learning algorithm will output information on the standard output stream System.out during learning.
java.lang.String	toString()

Method Detail

isValid

boolean isValid()

Test if the current settings are valid.

Returns:

true if the settings are valid, false otherwise

toString

java.lang.String toString()

Overrides:

toString in class java.lang.Object

copy

LearningSettings copy()

Make a deep copy of the learning settings.

Returns:

a deep copy of the current instance

getLearningAlgorithm

APIUtils.LEARNING_ALGORITHM getLearningAlgorithm()

Return a LEARNING_ALGORITHM indicating the learning algorithm to use

Returns:

the learning algorithm to use

isSupervised

boolean isSupervised()

Indicate if the current learning algorithm is supervised (i.e. it needs a target) or not.

Returns:

true if the learning algorithm is supervised, false if unsupervised. APIUtils.LEARNING_ALGORITHM.isSupervised()

getTargetNodeName

java.lang.String getTargetNodeName()

Return the name of the target node if the learning algorithm is supervised.

Returns:

the target node's name

getSmoothedProbabilityEstimation

double getSmoothedProbabilityEstimation()

Returns the weight of the non informative prior, i.e. the virtual case(s) spread uniformly across the joint probability distribution. This weight must be determined by taking into account the total weight of the database. It is not necessary a whole number. When 0 is returned, the learning process only uses the data from the database. The default value is 0.

Returns:

the weight associated to each virtual case

getStructuralCoefficient

double getStructuralCoefficient()

Returns the weight associated to the structural part used in the Minimum Description Length score. The higher its value, the stronger the probabilistic relations would have to be to result in a corresponding arc in the network. Conversely, the lower its value, the weaker the probabilistic relation can be while still being represented with an arc. By default, its value is set to a conservative value of 1. The default value is 1.

Returns:

the weight associated to the structural part of the MDL score

isVerbose

boolean isVerbose()

Tests if the learning algorithm will output information on the standard output stream System.out during learning.

Returns:

true if the learning algorithm prints output information