de.up.ling.irtg.sampling.rule_weighting

Class AutomatonWeighted

java.lang.Object

de.up.ling.irtg.sampling.rule_weighting.RegularizedKLRuleWeighting

de.up.ling.irtg.sampling.rule_weighting.AutomatonWeighted

All Implemented Interfaces:

RuleWeighting

public class AutomatonWeighted
extends RegularizedKLRuleWeighting

This class simply assigns each tree the weight of the multiplication of its rule weights. The underlying automaton is not changed in any way during sampling.

Author:

teichmann

Field Summary

Fields inherited from class de.up.ling.irtg.sampling.rule_weighting.RegularizedKLRuleWeighting

ALMOST_ZERO

Constructor Summary

Constructors

Constructor and Description
AutomatonWeighted(TreeAutomaton basis, int regularizationExponent, double regularizationDivisor, LearningRate rate) Creates a new instance that weights trees according to the rule weights from the given automaton and uses the automaton as the basis for sampling.

Method Summary

Modifier and Type	Method and Description
double	getLogTargetProbability(Tree<Rule> sample) Returns the unnormalized probability of the given tree in the target distribution.

Methods inherited from class de.up.ling.irtg.sampling.rule_weighting.RegularizedKLRuleWeighting

adapt, getAutomaton, getLogProbability, getLogProbability, getNumberOfStartStates, getRuleByNumber, getRuleNumber, getStartStateByNumber, getStartStateNumber, getStateStartLogProbability, prepareProbability, prepareStartProbability, reset

Methods inherited from class java.lang.Object

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

Constructor Detail

AutomatonWeighted

public AutomatonWeighted(TreeAutomaton basis,
                         int regularizationExponent,
                         double regularizationDivisor,
                         LearningRate rate)

Creates a new instance that weights trees according to the rule weights from the given automaton and uses the automaton as the basis for sampling. The adaption is implemented by the parent class.

Parameters:

basis -

regularizationExponent -

regularizationDivisor -

rate -

Method Detail

getLogTargetProbability

public double getLogTargetProbability(Tree<Rule> sample)

Description copied from interface: RuleWeighting

Returns the unnormalized probability of the given tree in the target distribution.

Returns: