Module org.chocosolver

Package org.chocosolver.solver.search.loop.move

Interface IMoveFactory

All Superinterfaces:

ISelf<Solver>

All Known Subinterfaces:

ISolver

All Known Implementing Classes:

Solver

public interface IMoveFactory
extends ISelf<Solver>

Interface to define how to explore the search space from a macro perspective (DFS, LDS, LNS, etc.)

Author:

Charles Prud'Homme, Jean-Guillaume Fages

Method Summary

Modifier and Type	Method	Description
default void	setDDS(int discrepancy)	Depth-bounded Discrepancy Search[1] algorithms with binary decisions
default void	setDFS()	Depth-First Search algorithm with binary decisions
default void	setGeometricalRestart(long base, double geometricalFactor, ICounter restartStrategyLimit, int restartLimit)	Build a geometrical restart strategy.
default void	setHBFS(double a, double b, long N)	Creates a move object based on: Hybrid Best-First Search[1] algorithms with binary decisions.
default void	setLDS(int discrepancy)	Limited Discrepancy Search[1] algorithms with binary decisions
default void	setLNS(INeighbor neighbor)	Creates a Move object based on Large Neighborhood Search.
default void	setLNS(INeighbor neighbor, ICounter restartCounter)	Creates a Move object based on Large Neighborhood Search.
default void	setLubyRestart(long scaleFactor, ICounter restartStrategyLimit, int restartLimit)	Branch a luby restart strategy to the model
default void	setRestartOnSolutions()	Creates a Move object that encapsulates the current move within a restart move.
default void	setRestarts(LongCriterion restartCriterion, org.chocosolver.cutoffseq.ICutoffStrategy restartStrategy, int restartsLimit)	Creates a Move object that encapsulates the current move within a restart move.

Methods inherited from interface org.chocosolver.solver.ISelf

ref

Method Detail

setDFS

default void setDFS()

Depth-First Search algorithm with binary decisions

setLDS

default void setLDS(int discrepancy)

Limited Discrepancy Search[1] algorithms with binary decisions

[1]:W.D. Harvey and M.L.Ginsberg, Limited Discrepancy Search, IJCAI-95.

Parameters:

discrepancy - the maximum discrepancy

setDDS

default void setDDS(int discrepancy)

Depth-bounded Discrepancy Search[1] algorithms with binary decisions

[1]:T. Walsh, Depth-bounded Discrepancy Search, IJCAI-97.

Parameters:

discrepancy - the maximum discrepancy

setHBFS

default void setHBFS(double a,
                     double b,
                     long N)

Creates a move object based on: Hybrid Best-First Search[1] algorithms with binary decisions.

[1]:D. Allouche, S. de Givry, G. Katsirelos, T. Schiex, M. Zytnicki, Anytime Hybrid Best-First Search with Tree Decomposition for Weighted CSP, CP-2015.

Parameters:

a - lower bound to limit the rate of redundantly propagated decisions

b - upper bound to limit the rate of redundantly propagated decisions.

N - backtrack limit for each DFS try, should be large enough to limit redundancy

setRestarts

default void setRestarts(LongCriterion restartCriterion,
                         org.chocosolver.cutoffseq.ICutoffStrategy restartStrategy,
                         int restartsLimit)

Creates a Move object that encapsulates the current move within a restart move. Every time the restartCriterion is met, a restart is done, the new restart limit is updated thanks to restartStrategy. There will be at most restartsLimit restarts.

Parameters:

restartCriterion - the restart criterion, that is, the condition which triggers a restart

restartStrategy - the way restart limit (evaluated in restartCriterion) is updated, that is, computes the next limit

restartsLimit - number of allowed restarts

setLubyRestart

default void setLubyRestart(long scaleFactor,
                            ICounter restartStrategyLimit,
                            int restartLimit)

Branch a luby restart strategy to the model

Parameters:

scaleFactor - scale factor

restartStrategyLimit - restart trigger

restartLimit - restart limits (limit of number of restarts)

setGeometricalRestart

default void setGeometricalRestart(long base,
                                   double geometricalFactor,
                                   ICounter restartStrategyLimit,
                                   int restartLimit)

Build a geometrical restart strategy. The restartStrategyLimit is a ICounter to observe. Once this counter reaches a limit, restart occurs and the next cutoff is computed. At step n, the next cutoff is computed with the following function : b*g^n.

Parameters:

base - the initial limit

geometricalFactor - geometrical factor

restartStrategyLimit - restart trigger

restartLimit - restart limits (limit of number of restarts)

setRestartOnSolutions

default void setRestartOnSolutions()

Creates a Move object that encapsulates the current move within a restart move. Every time a solution is found, a restart is done.

setLNS

default void setLNS(INeighbor neighbor,
                    ICounter restartCounter)

Creates a Move object based on Large Neighborhood Search. It encapsulates the current move within a LNS move. Anytime a solution is encountered, it is recorded and serves as a basis for the INeighbor. The INeighbor creates a fragment: selects variables to freeze/unfreeze wrt the last solution found. If a fragment cannot be extended to a solution, a new one is selected by restarting the search. If a fragment induces a search space which a too big to be entirely evaluated, restarting the search can be forced using the restartCriterion. A fast restart strategy is often a good choice.

Parameters:

neighbor - the neighbor for the LNS

restartCounter - the (fast) restart counter. Initial limit gives the frequency.

setLNS

default void setLNS(INeighbor neighbor)

Creates a Move object based on Large Neighborhood Search. It encapsulates the current move within a LNS move. Anytime a solution is encountered, it is recorded and serves as a basis for the INeighbor. The INeighbor creates a fragment: selects variables to freeze/unfreeze wrt the last solution found. If a fragment cannot be extended to a solution, a new one is selected by restarting the search.

Parameters:

neighbor - the neighbor for the LNS

See Also:

setLNS(INeighbor, ICounter)