The Branch and Prune Algorithm The Most Distant First Strategy The Depth and Most Distant First Strategy Experiments Conclusion and Future Work Search Strategies for an Anytime Usage of the Branch and Prune Algorithm R. Chenouard 1 , A. Goldsztejn 2 and C. Jermann 1 SWIM 2009 1 Lina UMR CNRS 6241, University of Nantes, France 2 CNRS, Lina UMR CNRS 6241, France LABORATOIRE D'INFORMATIQUE DE NANTES ATLANTIQUE lin R. Chenouard, A. Goldsztejn and C. Jermann Search Strategies for an Anytime Usage of the BPA 1/16
16
Embed
Search Strategies for an Anytime Usage of the Branch and Prune
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
The Branch and Prune AlgorithmThe Most Distant First Strategy
The Depth and Most Distant First StrategyExperiments
Conclusion and Future Work
Search Strategies for an Anytime Usageof the Branch and Prune Algorithm
R. Chenouard1, A. Goldsztejn2 and C. Jermann1
SWIM 2009
1 Lina UMR CNRS 6241, University of Nantes, France2 CNRS, Lina UMR CNRS 6241, France
LABORATOIRE D'INFORMATIQUEDE NANTES ATLANTIQUE
linR. Chenouard, A. Goldsztejn and C. Jermann Search Strategies for an Anytime Usage of the BPA 1/16
The Branch and Prune AlgorithmThe Most Distant First Strategy
The Depth and Most Distant First StrategyExperiments
Conclusion and Future Work
IntroductionExisting StrategiesIdeal Strategy for an Anytime Usage
The Branch and Prune AlgorithmAlternates search (branch) and filtering (prune).⇒ set of ε-boxes covering the solution set.
I Boxes to be processed : L
I Computed ε-boxes : S (can be post-processed)
Search IssueHow to choose the next box to process in L ?
R. Chenouard, A. Goldsztejn and C. Jermann Search Strategies for an Anytime Usage of the BPA 2/16
The Branch and Prune AlgorithmThe Most Distant First Strategy
The Depth and Most Distant First StrategyExperiments
Conclusion and Future Work
IntroductionExisting StrategiesIdeal Strategy for an Anytime Usage
Breadth-First Search
Principle of BFSSelects boxes in L following a FIFO strategy.
Main behaviorExplores uniformely the whole search space.
⇒ At time Ti no ε-boxes are computed.
R. Chenouard, A. Goldsztejn and C. Jermann Search Strategies for an Anytime Usage of the BPA 3/16
The Branch and Prune AlgorithmThe Most Distant First Strategy
The Depth and Most Distant First StrategyExperiments
Conclusion and Future Work
IntroductionExisting StrategiesIdeal Strategy for an Anytime Usage
Depth-First Search
Principle of DFSSelects boxes in L following a LIFO strategy.
Main behaviorDescends rapidly to ε-boxes then finds them by neighborhood.
⇒ At time Ti many similiar ε-boxes are computed.
R. Chenouard, A. Goldsztejn and C. Jermann Search Strategies for an Anytime Usage of the BPA 4/16
The Branch and Prune AlgorithmThe Most Distant First Strategy
The Depth and Most Distant First StrategyExperiments
Conclusion and Future Work
IntroductionExisting StrategiesIdeal Strategy for an Anytime Usage
Ideal Search Strategy for an Anytime UsagePrincipleSelects boxes in L mixing BFS and DFS principles.
Main behaviorExplores uniformely the whole search space and findwell-ditributed ε-boxes at early stage of the search.
R. Chenouard, A. Goldsztejn and C. Jermann Search Strategies for an Anytime Usage of the BPA 5/16
The Branch and Prune AlgorithmThe Most Distant First Strategy
The Depth and Most Distant First StrategyExperiments