Module org.chocosolver

Package org.chocosolver.solver.search.strategy.selectors.variables

Class ImpactBased

java.lang.Object

org.chocosolver.solver.search.strategy.strategy.AbstractStrategy<IntVar>

org.chocosolver.solver.search.strategy.selectors.variables.ImpactBased

All Implemented Interfaces:

ICause, IMonitorContradiction, IMonitorDownBranch, ISearchMonitor

public class ImpactBased
extends AbstractStrategy<IntVar>
implements IMonitorDownBranch, IMonitorContradiction, ICause

Implementation of the search described in: "Impact-Based Search Strategies for Constraint Programming", Philippe Refalo, CP2004.

Impacts of variables are first computed on initialisation, (only 3 values are tested for bounded variables); Ties are broken randomly.

Since:

21/09/12

Author:

Charles Prud'homme

Field Summary

Fields inherited from class org.chocosolver.solver.search.strategy.strategy.AbstractStrategy

vars

Constructor Summary

Constructors

Constructor	Description
ImpactBased(IntVar[] vars, boolean initOnly)
ImpactBased(IntVar[] ivariables, int alpha, int split, int nodeImpact, long seed, boolean initOnly)	Create an Impact-based search strategy with Node Impact strategy.

Method Summary

Modifier and Type	Method	Description
void	afterDownBranch(boolean left)	Action to perform after going down in the tree search
Decision<IntVar>	computeDecision(IntVar variable)	Computes a decision to be applied to variable var This method should be implemented in order to use search patterns
Decision<IntVar>	getDecision()	Provides access to the current decision in the strategy.
boolean	init()	Prepare this to be used in a search loop The initialization can detect inconsistency, in that case, it returns false
void	onContradiction(ContradictionException cex)
void	remove()	Remove the current strategy.
void	setInitialisationTimeLimit(long timeLimit)	Define a time limit on initialisation phase.
void	setReevaluationTimeLimit(long timeLimit)	Define a time limit on reevaluation phase (done every nodeImpact nodes).

Methods inherited from class org.chocosolver.solver.search.strategy.strategy.AbstractStrategy

getVariables, makeIntDecision, toString

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, wait, wait, wait

Methods inherited from interface org.chocosolver.solver.ICause

explain, forEachIntVar

Methods inherited from interface org.chocosolver.solver.search.loop.monitors.IMonitorDownBranch

beforeDownBranch

Constructor Detail

ImpactBased

public ImpactBased(IntVar[] ivariables,
                   int alpha,
                   int split,
                   int nodeImpact,
                   long seed,
                   boolean initOnly)

Create an Impact-based search strategy with Node Impact strategy.

"Impact-Based Search Strategies for Constraint Programming", Philippe Refalo, CP2004.

Parameters:

ivariables - variables of the problem (should be integers)

alpha - aging parameter

split - split parameter for subdomains computation

nodeImpact - force update of impacts every nodeImpact nodes. Set value to 0 to avoid using it.

seed - a seed for random

initOnly - only apply the initialisation phase, do not update impact thereafter

ImpactBased

public ImpactBased(IntVar[] vars,
                   boolean initOnly)

Method Detail

computeDecision

public Decision<IntVar> computeDecision(IntVar variable)

Description copied from class: AbstractStrategy

Computes a decision to be applied to variable var This method should be implemented in order to use search patterns

Overrides:

computeDecision in class AbstractStrategy<IntVar>

Parameters:

variable - a variable

Returns:

a decision to be applied to variable var

getDecision

public Decision<IntVar> getDecision()

Description copied from class: AbstractStrategy

Provides access to the current decision in the strategy. If there are no more decision to provide, it returns null.

Specified by:

getDecision in class AbstractStrategy<IntVar>

Returns:

the current decision

setInitialisationTimeLimit

public void setInitialisationTimeLimit(long timeLimit)

Define a time limit on initialisation phase. When the limit is reached, computing impacts on ROOT node stops and the search can start.

Parameters:

timeLimit - in milliseconds

setReevaluationTimeLimit

public void setReevaluationTimeLimit(long timeLimit)

Define a time limit on reevaluation phase (done every nodeImpact nodes). When the limit is reached, reevaluting impacts stops and the search can go on.

Parameters:

timeLimit - in milliseconds

init

public boolean init()

Description copied from class: AbstractStrategy

Prepare this to be used in a search loop The initialization can detect inconsistency, in that case, it returns false

Overrides:

init in class AbstractStrategy<IntVar>

remove

public void remove()

Description copied from class: AbstractStrategy

Remove the current strategy. This implies unplugging variable or search monitors.

Overrides:

remove in class AbstractStrategy<IntVar>

onContradiction

public void onContradiction(ContradictionException cex)

Specified by:

onContradiction in interface IMonitorContradiction

afterDownBranch

public void afterDownBranch(boolean left)

Description copied from interface: IMonitorDownBranch

Action to perform after going down in the tree search

Specified by:

afterDownBranch in interface IMonitorDownBranch

Parameters:

left - set to true to specify that this is a left branch