a guide to budgeted tree search
play

A Guide to Budgeted Tree Search Nathan R. Sturtevant University of - PowerPoint PPT Presentation

A Guide to Budgeted Tree Search Nathan R. Sturtevant University of Alberta Amii Fellow, CIFAR Chair Malte Helmert Universitt Basel Talk Overview Budgeted Tree Search (BTS) is a new algorithm with better worst-case guarantees than IDA*


  1. Nodes and f -costs • Exponential Search: • Find value in unbounded sorted array • Tree Search: • Find ( node expansions ) in ( f-costs ) • Nodes expansions non-decreasing with f -cost A Guide to Budgeted Tree Search 21

  2. How does BTS work? If the nodes in each iteration do not grow exponentially

  3. Budgeted search • Like exponential search on f-costs A Guide to Budgeted Tree Search 23

  4. Budgeted search • Like exponential search on f-costs f = ∞ f = 10 A Guide to Budgeted Tree Search 23

  5. Budgeted search • Like exponential search on f-costs f = 27.2 (2 × ) f = 30.7 (8 × ) f = ∞ f = 10 A Guide to Budgeted Tree Search 23

  6. Budgeted search • Like exponential search on f-costs f = 27.2 (2 × ) f = 30.7 (8 × ) f = ∞ f = 10 n i = 100 A Guide to Budgeted Tree Search 23

  7. Budgeted search • Like exponential search on f-costs f = 27.2 (2 × ) f = 30.7 (8 × ) f = ∞ f = 10 n i = 100 200 A Guide to Budgeted Tree Search 23

  8. Budgeted search • Like exponential search on f-costs f = 27.2 (2 × ) f = 30.7 (8 × ) f = ∞ f = 10 n i = 100 200 800 A Guide to Budgeted Tree Search 23

  9. Budgeted search • Like exponential search on f-costs f = 27.2 (2 × ) f = 30.7 (8 × ) f = ∞ f = 10 < 200 n i = 100 200 800 A Guide to Budgeted Tree Search 23

  10. Budgeted search • Like exponential search on f-costs f = 27.2 (2 × ) f = 30.7 (8 × ) f = ∞ f = 10 < 200 ≥ 800 n i = 100 200 800 A Guide to Budgeted Tree Search 23

  11. Budgeted search • Like exponential search on f-costs f = 27.2 (2 × ) f = 30.7 (8 × ) f = ∞ f = 10 < 200 ≥ 800 n i = 100 200 800 Budget: Stop when exceeded A Guide to Budgeted Tree Search 23

  12. Budgeted search • Like exponential search on f-costs f = 27.2 (2 × ) f = 30.7 (8 × ) f = ∞ f = 10 < 200 ≥ 800 n i = 100 200 800 Budget: Stop when exceeded A Guide to Budgeted Tree Search 23

  13. Budgeted search • Like exponential search on f-costs A Guide to Budgeted Tree Search 24

  14. Budgeted search • Like exponential search on f-costs f = 10 A Guide to Budgeted Tree Search 24

  15. Budgeted search • Like exponential search on f-costs f = 10 A Guide to Budgeted Tree Search 24

  16. Budgeted search • Like exponential search on f-costs f = ∞ f = 10 A Guide to Budgeted Tree Search 24

  17. Budgeted search • Like exponential search on f-costs f = 27.2 (2 × ) f = 30.7 (8 × ) f = ∞ f = 10 A Guide to Budgeted Tree Search 24

  18. Budgeted search • Like exponential search on f-costs f = 27.2 (2 × ) f = 30.7 (8 × ) f = ∞ f = 10 11 < A Guide to Budgeted Tree Search 24

  19. Budgeted search • Like exponential search on f-costs f = 27.2 (2 × ) f = 30.7 (8 × ) f = ∞ f = 10 11 12 < < A Guide to Budgeted Tree Search 24

  20. Budgeted search • Like exponential search on f-costs f = 27.2 (2 × ) f = 30.7 (8 × ) f = ∞ f = 10 14 11 12 < < < A Guide to Budgeted Tree Search 24

  21. Budgeted search • Like exponential search on f-costs f = 27.2 (2 × ) f = 30.7 (8 × ) f = ∞ f = 10 14 11 12 18 < < < < A Guide to Budgeted Tree Search 24

  22. Budgeted search • Like exponential search on f-costs f = 27.2 (2 × ) f = 30.7 (8 × ) f = ∞ f = 10 14 11 12 18 26 < < < < < A Guide to Budgeted Tree Search 24

  23. Budgeted search • Like exponential search on f-costs f = 27.2 (2 × ) f = 30.7 (8 × ) f = ∞ f = 10 14 11 12 18 26 42 < < < < < > A Guide to Budgeted Tree Search 24

  24. Budgeted search • Like exponential search on f-costs f = 27.2 (2 × ) f = 30.7 (8 × ) f = ∞ f = 10 14 34 11 12 18 26 42 < < < > < < > A Guide to Budgeted Tree Search 24

Recommend


More recommend