edu.illinois.cs.cogcomp.lbjava.infer

Class ILPInference

java.lang.Object

edu.illinois.cs.cogcomp.lbjava.infer.Inference

edu.illinois.cs.cogcomp.lbjava.infer.ILPInference

public class ILPInference
extends Inference

This class employs an ILPSolver to solve a constrained inference problem. When constructing an instance of this class in an LBJava source file, use one of the constructors that does not specify a head object. The generated code will fill in the head object automatically. The other constructor parameters are used to specify the ILP algorithm and enable textual output of ILP variable descriptions to STDOUT. Textual output of the ILP problem itself is controlled by the ILPSolver.

Author:

Nick Rizzolo

Field Summary

Fields

Modifier and Type	Field and Description
protected int	ID The identification number for this object, used in debug file names.
protected HashMap	indexMap Used during ILP constraint generation.
protected static int	nextID Keeps the next ID number for objects of this class.
protected int	returnIndex Used during ILP constraint generation.
protected boolean	returnNegation Used during ILP constraint generation.
protected edu.illinois.cs.cogcomp.infer.ilp.ILPSolver	solver The ILP algorithm.
protected boolean	tautology This flag is set if the constraints turn out to be true in all cases.
protected boolean	topLevel Used during ILP constraint generation.
protected int	verbosity Verbosity level.
static int	VERBOSITY_HIGH A possible setting for verbosity.
static int	VERBOSITY_LOW A possible setting for verbosity.
static int	VERBOSITY_NONE A possible setting for verbosity.

Fields inherited from class edu.illinois.cs.cogcomp.lbjava.infer.Inference

constraint, head, variables

Constructor Summary

Constructors

Constructor and Description
ILPInference() Don't use this constructor, since it doesn't set an ILP algorithm.
ILPInference(edu.illinois.cs.cogcomp.infer.ilp.ILPSolver a) Initializes the ILP algorithm, but not the head object.
ILPInference(edu.illinois.cs.cogcomp.infer.ilp.ILPSolver a, int v) Initializes the ILP algorithm, but not the head object.
ILPInference(Object h) Don't use this constructor, since it doesn't set an ILP algorithm.
ILPInference(Object h, edu.illinois.cs.cogcomp.infer.ilp.ILPSolver a) Sets the head object and the ILP algorithm.
ILPInference(Object h, edu.illinois.cs.cogcomp.infer.ilp.ILPSolver a, int v) Sets the head object and the ILP algorithm.

Method Summary

Modifier and Type	Method and Description
void	addConstraint(FirstOrderConstraint c) Adds a constraint to the inference.
protected int	createVariable(String d) Creates a new Boolean variable to represent the value of a subexpression of some constraint.
boolean	equals(Object o) Two Inference objects are equal when they have the same run-time type and store the same head object.
int	hashCode() Simply returns the head's hash code.
protected void	infer() Uses the provided ILP algorithm to solve the ILP proglem if it hasn't already been solved.
String	valueOf(Learner c, Object o) Retrieves the value of the specified variable as identified by the classifier and the object that produce that variable.
void	visit(PropositionalAtLeast c) Derived classes override this method to do some type of processing on constraints of the parameter's type.
void	visit(PropositionalConjunction c) Derived classes override this method to do some type of processing on constraints of the parameter's type.
void	visit(PropositionalConstant c) Derived classes override this method to do some type of processing on constraints of the parameter's type.
void	visit(PropositionalDisjunction c) Derived classes override this method to do some type of processing on constraints of the parameter's type.
void	visit(PropositionalDoubleImplication c) Derived classes override this method to do some type of processing on constraints of the parameter's type.
void	visit(PropositionalImplication c) Derived classes override this method to do some type of processing on constraints of the parameter's type.
void	visit(PropositionalNegation c) Derived classes override this method to do some type of processing on constraints of the parameter's type.
void	visit(PropositionalVariable c) Derived classes override this method to do some type of processing on constraints of the parameter's type.

Methods inherited from class edu.illinois.cs.cogcomp.lbjava.infer.Inference

exampleToString, getHead, getHeadFinderTypes, getHeadType, getNormalizer, getVariable, getVariable, satisfied, visit, visit, visit, visit, visit, visit, visit, visit, visit, visit, visit, visit, visit, visit, visitAll

Methods inherited from class java.lang.Object

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

Field Detail

VERBOSITY_NONE

public static final int VERBOSITY_NONE

A possible setting for verbosity.

See Also:

Constant Field Values

VERBOSITY_LOW

public static final int VERBOSITY_LOW

A possible setting for verbosity.

See Also:

Constant Field Values

VERBOSITY_HIGH

public static final int VERBOSITY_HIGH

A possible setting for verbosity.

See Also:

Constant Field Values

nextID

protected static int nextID

Keeps the next ID number for objects of this class.

ID

protected int ID

The identification number for this object, used in debug file names.

solver

protected edu.illinois.cs.cogcomp.infer.ilp.ILPSolver solver

The ILP algorithm.

tautology

protected boolean tautology

This flag is set if the constraints turn out to be true in all cases.

returnIndex

protected int returnIndex

Used during ILP constraint generation. When a propositional constraint finishes generating any ILP constraints that may be associated with it, it sets this variable to its own index.

returnNegation

protected boolean returnNegation

Used during ILP constraint generation. This flag is set iff the variable corresponding to returnIndex is negated in its current context.

indexMap

protected HashMap indexMap

Used during ILP constraint generation. This map associates each variable index with a representation of the expression whose value is represented by the variable. The keys associated with indexes of variables that were originally part of the inference problem are PropositionalVariable objects. The keys associated with indexes of temporary variables created during constraint translation are strings.

topLevel

protected boolean topLevel

Used during ILP constraint generation. Constraints are treated differently if they are part of another constraint expression than if they are a term in the top level conjunction.

verbosity

protected int verbosity

Verbosity level. VERBOSITY_NONE produces no incidental output. If set to VERBOSITY_LOW, only timing information is printed on STDOUT . If set to VERBOSITY_HIGH, information mapping the generated ILP variables to the first order variables they were generated from and their settings in the ILP problem's solution is printed to STDOUT.

Constructor Detail

ILPInference

public ILPInference()

Don't use this constructor, since it doesn't set an ILP algorithm.

ILPInference

public ILPInference(edu.illinois.cs.cogcomp.infer.ilp.ILPSolver a)

Initializes the ILP algorithm, but not the head object.

Parameters:

a - The ILP algorithm.

ILPInference

public ILPInference(edu.illinois.cs.cogcomp.infer.ilp.ILPSolver a,
                    int v)

Initializes the ILP algorithm, but not the head object.

Parameters:

a - The ILP algorithm.

v - Sets the value of verbosity.

ILPInference

public ILPInference(Object h)

Don't use this constructor, since it doesn't set an ILP algorithm.

ILPInference

public ILPInference(Object h,
                    edu.illinois.cs.cogcomp.infer.ilp.ILPSolver a)

Sets the head object and the ILP algorithm.

Parameters:

h - The head object.

a - The ILP algorithm.

ILPInference

public ILPInference(Object h,
                    edu.illinois.cs.cogcomp.infer.ilp.ILPSolver a,
                    int v)

Sets the head object and the ILP algorithm.

Parameters:

h - The head object.

a - The ILP algorithm.

v - Sets the value of verbosity.

Method Detail

addConstraint

public void addConstraint(FirstOrderConstraint c)

Adds a constraint to the inference.

Parameters:

c - The constraint to add.

infer

protected void infer()
              throws Exception

Uses the provided ILP algorithm to solve the ILP proglem if it hasn't already been solved.

Specified by:

infer in class Inference

Throws:

Exception

valueOf

public String valueOf(Learner c,
                      Object o)
               throws Exception

Retrieves the value of the specified variable as identified by the classifier and the object that produce that variable.

Specified by:

valueOf in class Inference

Parameters:

c - The classifier producing the variable.

o - The object from which the variable is produced.

Returns:

The current value of the requested variable. If the variable does not exist in this inference, the result of the Learner's discreteValue(Object) method applied to the Object is returned.

Throws:

Exception

equals

public boolean equals(Object o)

Two Inference objects are equal when they have the same run-time type and store the same head object. I.e., the == operator must return true when comparing the two head objects for this method to return true.

Overrides:

equals in class Object

Parameters:

o - The object to compare to this object.

Returns:

true iff this object equals the argument object as defined above.

hashCode

public int hashCode()

Simply returns the head's hash code.

Overrides:

hashCode in class Object

See Also:

Object.hashCode()

createVariable

protected int createVariable(String d)

Creates a new Boolean variable to represent the value of a subexpression of some constraint.

Parameters:

d - A textual description of the subexpression whose value is represented by the new variable.

Returns:

The index of the new variable.

visit

public void visit(PropositionalDoubleImplication c)

Derived classes override this method to do some type of processing on constraints of the parameter's type.

Overrides:

visit in class Inference

Parameters:

c - The constraint to process.

visit

public void visit(PropositionalImplication c)

Derived classes override this method to do some type of processing on constraints of the parameter's type.

Overrides:

visit in class Inference

Parameters:

c - The constraint to process.

visit

public void visit(PropositionalConjunction c)

Derived classes override this method to do some type of processing on constraints of the parameter's type.

Overrides:

visit in class Inference

Parameters:

c - The constraint to process.

visit

public void visit(PropositionalDisjunction c)

Derived classes override this method to do some type of processing on constraints of the parameter's type.

Overrides:

visit in class Inference

Parameters:

c - The constraint to process.

visit

public void visit(PropositionalAtLeast c)

Derived classes override this method to do some type of processing on constraints of the parameter's type.

Overrides:

visit in class Inference

Parameters:

c - The constraint to process.

visit

public void visit(PropositionalNegation c)

Derived classes override this method to do some type of processing on constraints of the parameter's type.

Overrides:

visit in class Inference

Parameters:

c - The constraint to process.

visit

public void visit(PropositionalVariable c)

Derived classes override this method to do some type of processing on constraints of the parameter's type.

Overrides:

visit in class Inference

Parameters:

c - The constraint to process.

visit

public void visit(PropositionalConstant c)

Derived classes override this method to do some type of processing on constraints of the parameter's type.

Overrides:

visit in class Inference

Parameters:

c - The constraint to process.