edu.stanford.nlp.stats

Interface ConjugatePrior<T extends ProbabilityDistribution<E>,E>

Type Parameters:

T -

E -

All Superinterfaces:

ProbabilityDistribution<T>, java.io.Serializable

All Known Implementing Classes:

Dirichlet

public interface ConjugatePrior<T extends ProbabilityDistribution<E>,E>
extends ProbabilityDistribution<T>

An interface for classes which are probability distributions over other probability distributions. This class could also represent non-conjugate priors, but we thought this name would be more representative of the common case.

Author:

grenager, jrfinkel

Method Summary

Modifier and Type	Method and Description
ConjugatePrior<T,E>	getPosteriorDistribution(Counter<E> evidence) Gets the ConjugatePrior which results from conditioning on all of these evidence.
double	getPosteriorPredictiveLogProbability(Counter<E> evidence, E observation)
double	getPosteriorPredictiveProbability(Counter<E> evidence, E observation) Gets the posterior probability of the observation, after conditioning on all of the evidence.
double	getPredictiveLogProbability(E observation)
double	getPredictiveProbability(E observation) Marginalizes over all possible likelihood distributions to give the marginal probability of the observation.

Methods inherited from interface edu.stanford.nlp.stats.ProbabilityDistribution

drawSample, logProbabilityOf, probabilityOf

Method Detail

getPredictiveProbability

double getPredictiveProbability(E observation)

Marginalizes over all possible likelihood distributions to give the marginal probability of the observation.

getPredictiveLogProbability

double getPredictiveLogProbability(E observation)

getPosteriorPredictiveProbability

double getPosteriorPredictiveProbability(Counter<E> evidence,
                                         E observation)

Gets the posterior probability of the observation, after conditioning on all of the evidence. Marginalizes over all possible likelihood distributions.

getPosteriorPredictiveLogProbability

double getPosteriorPredictiveLogProbability(Counter<E> evidence,
                                            E observation)

getPosteriorDistribution

ConjugatePrior<T,E> getPosteriorDistribution(Counter<E> evidence)

Gets the ConjugatePrior which results from conditioning on all of these evidence.