1 / 19

A Little Respect (for the Role of Common Components in Heuristic Search)

A Little Respect (for the Role of Common Components in Heuristic Search). Stephen Chen York University. Heuristic Search. Search Operators Control Strategies. An Obvious Failing. Point search operators do not have historical information! Most research focuses on the control strategies.

wmichele
Download Presentation

A Little Respect (for the Role of Common Components in Heuristic Search)

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. A Little Respect (for the Role of Common Components in Heuristic Search) Stephen Chen York University

  2. Heuristic Search • Search Operators • Control Strategies IFIP WCC 2006

  3. An Obvious Failing • Point search operators do not have historical information! • Most research focuses on the control strategies IFIP WCC 2006

  4. Search Operator Strategies • What can be improved? • What is worth keeping? • Most search operators focus on what they change, not what they keep IFIP WCC 2006

  5. What Should Search Operators Keep? • Would like to keep “above-average” components of existing solutions and change their “below-average” components • Use Commonality Hypothesis to identify “above-average” components IFIP WCC 2006

  6. The Commonality Hypothesis • Schemata common to above-average solutions are above average • Search operators should keep common components and change uncommon components IFIP WCC 2006

  7. A Simple Example Parent 1 1 1 1 0 0 1 0 1 1 1 Parent 2 1 0 1 1 0 1 1 0 1 1 Common 1 1 0 1 1 1 Uncommon 1 1 0 0 1 Uncommon 2 0 1 1 0 IFIP WCC 2006

  8. TSP Nearest Neighbour Tours Tour 1 Tour 2 IFIP WCC 2006

  9. Common and Uncommon Components Common Uncommon IFIP WCC 2006

  10. Random Local Minima IFIP WCC 2006

  11. Simulated Annealing IFIP WCC 2006

  12. Summary • Random search space, random search strategy • All the intelligence of simulated annealing is in the control strategy IFIP WCC 2006

  13. Distribution of Local Minima (Boese) Length Average distance from other local minima IFIP WCC 2006

  14. “Big-Valley” – Globally Convex IFIP WCC 2006

  15. “Directed” Simulated Annealing IFIP WCC 2006

  16. Coordination with Common Components IFIP WCC 2006

  17. TSP Results IFIP WCC 2006

  18. Summary • Globally convex search space, “respectful” search operator • Add some intelligence to the search operator IFIP WCC 2006

  19. Conclusion • The preservation of common components can improve the performance of heuristic search procedures in globally convex search spaces IFIP WCC 2006

More Related