cc.mallet.grmm.types
Class PottsTableFactor

java.lang.Object
  cc.mallet.grmm.types.AbstractFactor
      cc.mallet.grmm.types.PottsTableFactor

All Implemented Interfaces:

Factor, ParameterizedFactor, java.io.Serializable, java.lang.Cloneable

public class PottsTableFactor
extends AbstractFactor
implements ParameterizedFactor

A factor over a continuous variable alpha and discrete variables x such that phi(x|alpha) is Potts. That is, for fixed alpha, phi(x) = 1 if all x are equal, and exp^{-alpha} otherwise. $Id: PottsTableFactor.java,v 1.1 2007/10/22 21:37:44 mccallum Exp $

See Also:

Serialized Form

Field Summary

Fields inherited from class cc.mallet.grmm.types.AbstractFactor

vars

Constructor Summary

PottsTableFactor(Variable x1, Variable x2, Variable alpha)

PottsTableFactor(VarSet xs, Variable alpha)

Method Summary

boolean	almostEquals(Factor p, double epsilon)
java.lang.String	dumpToString()
Factor	duplicate()
boolean	equals(java.lang.Object o)
protected Factor	extractMaxInternal(VarSet varSet)
int	hashCode()
boolean	isNaN()
double	logValue(AssignmentIterator it)
protected double	lookupValueInternal(int i)
protected Factor	marginalizeInternal(VarSet varsToKeep)
Factor	normalize() Multiplies this potential by a constant such that it sums to 1.
Assignment	sample(Randoms r) Return an assignment sampled from this factor, interpreting it as an unnormalized probability distribution.
double	secondDerivative(Factor q, Variable param, Assignment theta)
Factor	slice(Assignment assn)
double	sumGradLog(Factor q, Variable param, Assignment theta) Computes the expected derivative of the log factor value.
double	value(AssignmentIterator it) Returns the probability of an assignment to these variables.

Methods inherited from class cc.mallet.grmm.types.AbstractFactor

almostEquals, argmax, assignmentIterator, asTable, containsVar, delogify, divideBy, entropy, exponentiate, extractMax, extractMax, extractMax, getVariable, isInLogSpace, log, logify, logValue, logValue, marginalize, marginalize, marginalize, marginalizeOut, marginalizeOut, multiply, multiplyBy, phi, prettyOutputString, setVarSet, sum, value, varSet

Methods inherited from class java.lang.Object

clone, finalize, getClass, notify, notifyAll, toString, wait, wait, wait

Methods inherited from interface cc.mallet.grmm.types.Factor

almostEquals, argmax, assignmentIterator, asTable, containsVar, divideBy, entropy, exponentiate, extractMax, extractMax, extractMax, getVariable, logValue, logValue, marginalize, marginalize, marginalize, marginalizeOut, marginalizeOut, multiply, multiplyBy, prettyOutputString, sum, value, varSet

Constructor Detail

PottsTableFactor

public PottsTableFactor(VarSet xs,
                        Variable alpha)

PottsTableFactor

public PottsTableFactor(Variable x1,
                        Variable x2,
                        Variable alpha)

Method Detail

extractMaxInternal

protected Factor extractMaxInternal(VarSet varSet)

Specified by:

extractMaxInternal in class AbstractFactor

lookupValueInternal

protected double lookupValueInternal(int i)

Specified by:

lookupValueInternal in class AbstractFactor

marginalizeInternal

protected Factor marginalizeInternal(VarSet varsToKeep)

Specified by:

marginalizeInternal in class AbstractFactor

value

public double value(AssignmentIterator it)

Description copied from interface: Factor

Returns the probability of an assignment to these variables. The assignment used is the curret assignment from the given AssignmentIterator.

This can be used to do things like

         DiscretePotential phi = createMyPtl ();
         for (AssignmentIterator it = phi.assignmentIterator; it.hasNext(); it.advance()) {
           double val = ptl.phi (it);
                 // do something with val
   }
  

This is equivalent to creating an assignment object explicitly using (Assignment) it.next(), but can be much faster.

Specified by:

value in interface Factor

Overrides:

value in class AbstractFactor

normalize

public Factor normalize()

Description copied from interface: Factor

Multiplies this potential by a constant such that it sums to 1. Destructive; returns this factor.

Specified by:

normalize in interface Factor

sample

public Assignment sample(Randoms r)

Description copied from interface: Factor

Return an assignment sampled from this factor, interpreting it as an unnormalized probability distribution.

Specified by:

sample in interface Factor

Overrides:

sample in class AbstractFactor

logValue

public double logValue(AssignmentIterator it)

Specified by:

logValue in interface Factor

Overrides:

logValue in class AbstractFactor

slice

public Factor slice(Assignment assn)

Specified by:

slice in interface Factor

dumpToString

public java.lang.String dumpToString()

Specified by:

dumpToString in interface Factor

sumGradLog

public double sumGradLog(Factor q,
                         Variable param,
                         Assignment theta)

Description copied from interface: ParameterizedFactor

Computes the expected derivative of the log factor value. That is,

sum_{y} q(y) dlog f(y) / d theta
,
 where y are the outcomes of the discrete varables in the factor,
  f(y) is the factor value, and theta is the vector of continuous variables
  in the factor.  q is a user-specified distribution to take the expectation
  with respect to.
  

  The factor q specifies with variables to sum over.  The summation will be over
  all the variables in q.varSet(), and the rest of the variables will be used

  



Specified by:
sumGradLog in interface ParameterizedFactor


Parameters:
q - Distribution to take with respect to (need not be normalized).
    q.varSet() should be all of the variables of this factor, except for one continuous variable
param - Parameter to take gradient with respect to.
Returns:
The expected gradient

secondDerivative

public double secondDerivative(Factor q,
                               Variable param,
                               Assignment theta)

duplicate

public Factor duplicate()

Specified by:

duplicate in interface Factor

isNaN

public boolean isNaN()

Specified by:

isNaN in interface Factor

almostEquals

public boolean almostEquals(Factor p,
                            double epsilon)

Specified by:

almostEquals in interface Factor

equals

public boolean equals(java.lang.Object o)

Overrides:

equals in class java.lang.Object

hashCode

public int hashCode()

Overrides:

hashCode in class java.lang.Object