parallel algorithms for graphs on a very large number of
play

Parallel Algorithms for Graphs on a Very Large Number of Nodes - PowerPoint PPT Presentation

Mar 12, 2024 •137 likes •1.17k views

Parallel Algorithms for Graphs on a Very Large Number of Nodes Krzysztof Onak IBM T.J. Watson Research Center Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 1 / 26 Outline Model of Computation 1 Sample

  1. N 1 − Ω( 1 ) Space in O ( 1 ) Rounds? • Unlikely to be possible in general Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 10 / 26

  2. N 1 − Ω( 1 ) Space in O ( 1 ) Rounds? • Unlikely to be possible in general • Can reduce from Sparse Connectivity: Do edges span a connected graph? Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 10 / 26

  3. N 1 − Ω( 1 ) Space in O ( 1 ) Rounds? • Unlikely to be possible in general • Can reduce from Sparse Connectivity: Do edges span a connected graph? • Conjecture: superconstant number of rounds with N 1 − Ω( 1 ) memory Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 10 / 26

  4. N 1 − Ω( 1 ) Space in O ( 1 ) Rounds? • Unlikely to be possible in general • Can reduce from Sparse Connectivity: Do edges span a connected graph? • Conjecture: superconstant number of rounds with N 1 − Ω( 1 ) memory • Is this instance hard? vs. Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 10 / 26

  5. N 1 − Ω( 1 ) Space in O ( 1 ) Rounds? • Unlikely to be possible in general • Can reduce from Sparse Connectivity: Do edges span a connected graph? • Conjecture: superconstant number of rounds with N 1 − Ω( 1 ) memory • Is this instance hard? (solvable in O ( log N ) rounds) vs. Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 10 / 26

  6. N 1 − Ω( 1 ) Space in O ( 1 ) Rounds? • Unlikely to be possible in general • Can reduce from Sparse Connectivity: Do edges span a connected graph? • Conjecture: superconstant number of rounds with N 1 − Ω( 1 ) memory • Is this instance hard? (solvable in O ( log N ) rounds) vs. • Reduction: connect select vertex to all vertices with heavy edges Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 10 / 26

  7. N 1 − Ω( 1 ) Space in O ( 1 ) Rounds? • Unlikely to be possible in general • Can reduce from Sparse Connectivity: Do edges span a connected graph? • Conjecture: superconstant number of rounds with N 1 − Ω( 1 ) memory • Is this instance hard? (solvable in O ( log N ) rounds) vs. • Reduction: connect select vertex to all vertices with heavy edges • This talk: algorithms with O ( N ǫ ) space per machine Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 10 / 26

  8. Outline Model of Computation 1 Sample Algorithms and Their Limitations 2 Efficiently Estimating MST Weight 3 Computing MST in Geometric Setting 4 Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 11 / 26

  9. Result [Ł ˛ acki, M ˛ adry, Mitrovi´ c, O., Sankowski] • Input: M edges, weights in { 1 , 2 , . . . , W } (#nodes N ≤ #edges M ) Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 12 / 26

  10. Result [Ł ˛ acki, M ˛ adry, Mitrovi´ c, O., Sankowski] • Input: M edges, weights in { 1 , 2 , . . . , W } (#nodes N ≤ #edges M ) • Algorithm: • Computes ( 1 + ǫ ) -approximation to MST weight Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 12 / 26

  11. Result [Ł ˛ acki, M ˛ adry, Mitrovi´ c, O., Sankowski] • Input: M edges, weights in { 1 , 2 , . . . , W } (#nodes N ≤ #edges M ) • Algorithm: • Computes ( 1 + ǫ ) -approximation to MST weight • Space per machine: � � 2 � M m + N � W for M / m = M Ω( 1 ) O m · ǫ Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 12 / 26

  12. Result [Ł ˛ acki, M ˛ adry, Mitrovi´ c, O., Sankowski] • Input: M edges, weights in { 1 , 2 , . . . , W } (#nodes N ≤ #edges M ) • Algorithm: • Computes ( 1 + ǫ ) -approximation to MST weight • Space per machine: � � 2 � M m + N � W for M / m = M Ω( 1 ) O m · ǫ • Number of rounds: O ( log ( W /ǫ )) Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 12 / 26

  13. Result [Ł ˛ acki, M ˛ adry, Mitrovi´ c, O., Sankowski] • Input: M edges, weights in { 1 , 2 , . . . , W } (#nodes N ≤ #edges M ) • Algorithm: • Computes ( 1 + ǫ ) -approximation to MST weight • Space per machine: � � 2 � M m + N � W for M / m = M Ω( 1 ) O m · ǫ • Number of rounds: O ( log ( W /ǫ )) • Note: No dependence on W would disprove Sparse Connectivity Conjecture Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 12 / 26

  14. Approach Use techniques of Chazelle, Rubinfeld, Trevisan (2005) Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 13 / 26

  15. Approach Use techniques of Chazelle, Rubinfeld, Trevisan (2005) : • G i = graph restricted to edges of weight < i Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 13 / 26

  16. Approach Use techniques of Chazelle, Rubinfeld, Trevisan (2005) : • G i = graph restricted to edges of weight < i • T i = #connected components in G i Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 13 / 26

  17. Approach Use techniques of Chazelle, Rubinfeld, Trevisan (2005) : • G i = graph restricted to edges of weight < i • T i = #connected components in G i • Number of edges of weight ≥ i in MST = T i − 1 Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 13 / 26

  18. Approach Use techniques of Chazelle, Rubinfeld, Trevisan (2005) : • G i = graph restricted to edges of weight < i • T i = #connected components in G i • Number of edges of weight ≥ i in MST = T i − 1 W � ⇒ weight ( MST ) = ( T i − 1 ) i = 1 Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 13 / 26

  19. Approach Use techniques of Chazelle, Rubinfeld, Trevisan (2005) : • G i = graph restricted to edges of weight < i • T i = #connected components in G i • Number of edges of weight ≥ i in MST = T i − 1 W � ⇒ weight ( MST ) = ( T i − 1 ) i = 1 • C i ( v ) = size of the component of v in G i � T i = 1 / C i ( v ) v Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 13 / 26

  20. Approach Use techniques of Chazelle, Rubinfeld, Trevisan (2005) : • G i = graph restricted to edges of weight < i • T i = #connected components in G i • Number of edges of weight ≥ i in MST = T i − 1 W � ⇒ weight ( MST ) = ( T i − 1 ) i = 1 • C i ( v ) = size of the component of v in G i � T i = 1 / C i ( v ) v • Good approximation: • Compute sizes of small components • Replace 1 / C i ( v ) with 0 if C i ( v ) ≥ W /ǫ Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 13 / 26

  21. Implementation • Reachability sets R v for each node v : • Set of W /ǫ nodes accessible via cheapest edges Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 14 / 26

  22. Implementation • Reachability sets R v for each node v : • Set of W /ǫ nodes accessible via cheapest edges • Initially: collect cheapest incident edges Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 14 / 26

  23. Implementation • Reachability sets R v for each node v : • Set of W /ǫ nodes accessible via cheapest edges • Initially: collect cheapest incident edges • Repeat O ( log ( W /ǫ )) times: Ask nodes u on R v for their R u and update Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 14 / 26

  24. Implementation • Reachability sets R v for each node v : • Set of W /ǫ nodes accessible via cheapest edges • Initially: collect cheapest incident edges • Repeat O ( log ( W /ǫ )) times: Ask nodes u on R v for their R u and update • O ( log ( W /ǫ )) updates suffice to explore useful nodes up to distance W /ǫ Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 14 / 26

  25. Implementation • Reachability sets R v for each node v : • Set of W /ǫ nodes accessible via cheapest edges • Initially: collect cheapest incident edges • Repeat O ( log ( W /ǫ )) times: Ask nodes u on R v for their R u and update • O ( log ( W /ǫ )) updates suffice to explore useful nodes up to distance W /ǫ • Use QuickSort-like sorting algorithm of Goodrich, Sitchinava, Zhang (2011) to organize communication Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 14 / 26

  26. Outline Model of Computation 1 Sample Algorithms and Their Limitations 2 Efficiently Estimating MST Weight 3 Computing MST in Geometric Setting 4 Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 15 / 26

  27. Geometric Setting Input: set of points in low dimensional metric space Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 16 / 26

  28. Geometric Setting Input: set of points in low dimensional metric space 7 9 14 8 11 10 • Points induce a weighted graph Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 16 / 26

  29. Geometric Setting Input: set of points in low dimensional metric space 7 9 14 8 11 10 • Points induce a weighted graph • Graph problems to consider: • Minimum Spanning Tree • Earth Mover Distance • Transportation Problem • Travelling Salesman Problem • . . . Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 16 / 26

  30. Result [Andoni, Nikolov, O., Yaroslavtsev 2014] • Input: N points in low dimensional metric space • Example: R 2 • Generalizes to bounded doubling dimension Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 17 / 26

  31. Result [Andoni, Nikolov, O., Yaroslavtsev 2014] • Input: N points in low dimensional metric space • Example: R 2 • Generalizes to bounded doubling dimension • Algorithm: • Computes ( 1 + ǫ ) -approximate MST Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 17 / 26

  32. Result [Andoni, Nikolov, O., Yaroslavtsev 2014] • Input: N points in low dimensional metric space • Example: R 2 • Generalizes to bounded doubling dimension • Algorithm: • Computes ( 1 + ǫ ) -approximate MST • Space per machine: roughly O ( N / m ) (as long as it fits subproblems) Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 17 / 26

  33. Result [Andoni, Nikolov, O., Yaroslavtsev 2014] • Input: N points in low dimensional metric space • Example: R 2 • Generalizes to bounded doubling dimension • Algorithm: • Computes ( 1 + ǫ ) -approximate MST • Space per machine: roughly O ( N / m ) (as long as it fits subproblems) • Number of rounds: O ( 1 ) Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 17 / 26

  34. Result [Andoni, Nikolov, O., Yaroslavtsev 2014] • Input: N points in low dimensional metric space • Example: R 2 • Generalizes to bounded doubling dimension • Algorithm: • Computes ( 1 + ǫ ) -approximate MST • Space per machine: roughly O ( N / m ) (as long as it fits subproblems) • Number of rounds: O ( 1 ) • Running time: near-linear Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 17 / 26

  35. Random Gridding We reuse the Arora-Mitchell approach: Apply a randomly shifted grid Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 18 / 26

  36. Random Gridding We reuse the Arora-Mitchell approach: Apply a randomly shifted grid Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 18 / 26

  37. Random Gridding We reuse the Arora-Mitchell approach: Apply a randomly shifted grid Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 18 / 26

  38. Random Gridding We reuse the Arora-Mitchell approach: Apply a randomly shifted grid Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 18 / 26

  39. Random Gridding We reuse the Arora-Mitchell approach: Apply a randomly shifted grid Key property: cell of side ∆ separates points x and y w.p. O ( 1 ) · ρ ( x , y ) ∆ Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 18 / 26

  40. Using Random Gridding Typical usage: Recursive dynamic program for approximately solving problem Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 19 / 26

  41. Using Random Gridding Typical usage: Recursive dynamic program for approximately solving problem Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 19 / 26

  42. Using Random Gridding Typical usage: Recursive dynamic program for approximately solving problem Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 19 / 26

  43. Using Random Gridding Typical usage: Recursive dynamic program for approximately solving problem Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 19 / 26

  44. Using Random Gridding Typical usage: Recursive dynamic program for approximately solving problem Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 19 / 26

  45. Using Random Gridding Typical usage: Recursive dynamic program for approximately solving problem Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 19 / 26

  46. Using Random Gridding Typical usage: Recursive dynamic program for approximately solving problem Can partially isolate what happens inside a cell Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 19 / 26

  47. Our Algorithm • Connect points closer than ǫ · diam ( S ) arbitrarily 100 · N Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 20 / 26

  48. Our Algorithm • Connect points closer than ǫ · diam ( S ) arbitrarily 100 · N • Sub-solution for cell of side ∆ : ǫ 2 ∆ -covering with induced components Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 20 / 26

  49. Our Algorithm • Connect points closer than ǫ · diam ( S ) arbitrarily 100 · N • Sub-solution for cell of side ∆ : ǫ 2 ∆ -covering with induced components • Combining sub-solutions: Truncated version of Kruskal’s algorithm Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 20 / 26

  50. Our Algorithm • Connect points closer than ǫ · diam ( S ) arbitrarily 100 · N • Sub-solution for cell of side ∆ : ǫ 2 ∆ -covering with induced components • Combining sub-solutions: Truncated version of Kruskal’s algorithm 1. Find two closest clusters Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 20 / 26

  51. Our Algorithm • Connect points closer than ǫ · diam ( S ) arbitrarily 100 · N • Sub-solution for cell of side ∆ : ǫ 2 ∆ -covering with induced components • Combining sub-solutions: Truncated version of Kruskal’s algorithm 1. Find two closest clusters 2. If their distance less than ǫ ∆ , connect them Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 20 / 26

  52. Our Algorithm • Connect points closer than ǫ · diam ( S ) arbitrarily 100 · N • Sub-solution for cell of side ∆ : ǫ 2 ∆ -covering with induced components • Combining sub-solutions: Truncated version of Kruskal’s algorithm 1. Find two closest clusters 2. If their distance less than ǫ ∆ , connect them Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 20 / 26

  53. Our Algorithm • Connect points closer than ǫ · diam ( S ) arbitrarily 100 · N • Sub-solution for cell of side ∆ : ǫ 2 ∆ -covering with induced components • Combining sub-solutions: Truncated version of Kruskal’s algorithm 1. Find two closest clusters 2. If their distance less than ǫ ∆ , connect them and repeat Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 20 / 26

  54. Our Algorithm • Connect points closer than ǫ · diam ( S ) arbitrarily 100 · N • Sub-solution for cell of side ∆ : ǫ 2 ∆ -covering with induced components • Combining sub-solutions: Truncated version of Kruskal’s algorithm 1. Find two closest clusters 2. If their distance less than ǫ ∆ , connect them and repeat • Pass up ǫ 2 ∆ -covering with information about connected components Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 20 / 26

  55. Our Algorithm • Connect points closer than ǫ · diam ( S ) arbitrarily 100 · N • Sub-solution for cell of side ∆ : ǫ 2 ∆ -covering with induced components • Combining sub-solutions: Truncated version of Kruskal’s algorithm 1. Find two closest clusters 2. If their distance less than ǫ ∆ , connect them and repeat • Pass up ǫ 2 ∆ -covering with information about connected components Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 20 / 26

  56. Our Algorithm • Connect points closer than ǫ · diam ( S ) arbitrarily 100 · N • Sub-solution for cell of side ∆ : ǫ 2 ∆ -covering with induced components • Combining sub-solutions: Truncated version of Kruskal’s algorithm 1. Find two closest clusters 2. If their distance less than ǫ ∆ , connect them and repeat • Pass up ǫ 2 ∆ -covering with information about connected components Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 20 / 26

  57. Our Algorithm • Connect points closer than ǫ · diam ( S ) arbitrarily 100 · N • Sub-solution for cell of side ∆ : ǫ 2 ∆ -covering with induced components • Combining sub-solutions: Truncated version of Kruskal’s algorithm 1. Find two closest clusters 2. If their distance less than ǫ ∆ , connect them and repeat • Pass up ǫ 2 ∆ -covering with information about connected components • Expected cost of solution: optimum · ( 1 + ǫ · #levels ) Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 20 / 26

  58. Select Implementation Details • Merge N Ω( 1 ) × N Ω( 1 ) cells at once Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 21 / 26

  59. Select Implementation Details • Merge N Ω( 1 ) × N Ω( 1 ) cells at once • Sub-solutions for all subcells should fit on a single machine Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 21 / 26

  60. Select Implementation Details • Merge N Ω( 1 ) × N Ω( 1 ) cells at once • Sub-solutions for all subcells should fit on a single machine • Use sorting [Goodrich, Sitchinava, Zhang 2011] for grouping points and subcells that are close Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 21 / 26

  61. Select Implementation Details • Merge N Ω( 1 ) × N Ω( 1 ) cells at once • Sub-solutions for all subcells should fit on a single machine • Use sorting [Goodrich, Sitchinava, Zhang 2011] for grouping points and subcells that are close • Near-linear time: • Relax Kruskal’s algorithm • Efficient nearest neighbor data structure [Krauthgamer, Lee 2004], [Cole, Gottlieb 2006] Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 21 / 26

  62. Lower Bounds for MST • Natural questions to ask: • Can generalize to unbounded dimension? • Can compute exact solution? Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 22 / 26

  63. Lower Bounds for MST • Natural questions to ask: • Can generalize to unbounded dimension? • Can compute exact solution? • Query complexity: • Model: distance queries • Our algorithm can be adapted to arbitrary bounded doubling dimensional metric in this model • Lower bound: N Ω( 1 ) rounds Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 22 / 26

  64. Lower Bounds for MST • Natural questions to ask: • Can generalize to unbounded dimension? • Can compute exact solution? • Query complexity: • Model: distance queries • Our algorithm can be adapted to arbitrary bounded doubling dimensional metric in this model • Lower bound: N Ω( 1 ) rounds • We give a conditional lower bound based on Sparse Connectivity Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 22 / 26

  65. Reduction In constant number of rounds: Computing exact MST in ℓ d ∞ for d = 100 log N ⇒ deciding Sparse Connectivity Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 23 / 26

  66. Reduction In constant number of rounds: Computing exact MST in ℓ d ∞ for d = 100 log N ⇒ deciding Sparse Connectivity Construction: • For each vertex, pick a random vector v i in {− 1 , + 1 } d • For each edge e = ( i , j ) , add point f ( e ) = v i + v j Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 23 / 26

  67. Reduction In constant number of rounds: Computing exact MST in ℓ d ∞ for d = 100 log N ⇒ deciding Sparse Connectivity Construction: • For each vertex, pick a random vector v i in {− 1 , + 1 } d • For each edge e = ( i , j ) , add point f ( e ) = v i + v j Distances (whp.): • Adjacent edges: � f ( e ) − f ( e ′ ) � ∞ ≤ 2 • Non-adjacent edges: � f ( e ) − f ( e ′ ) � ∞ = 4 Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 23 / 26

  68. Reduction In constant number of rounds: Computing exact MST in ℓ d ∞ for d = 100 log N ⇒ deciding Sparse Connectivity Construction: • For each vertex, pick a random vector v i in {− 1 , + 1 } d • For each edge e = ( i , j ) , add point f ( e ) = v i + v j Distances (whp.): • Adjacent edges: � f ( e ) − f ( e ′ ) � ∞ ≤ 2 • Non-adjacent edges: � f ( e ) − f ( e ′ ) � ∞ = 4 MST weight: • Connected: ≤ 2 ( M − 1 ) • Not connected: ≥ 2 M Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 23 / 26

  69. Other Results [Andoni, Nikolov, O., Yaroslavtsev 2014] • Algorithm for approximating Earth-Mover Distance • A new way of partitioning the instance into subproblems • Resolves an open question of Sharathkumar & Agarwal (2012) about the transportation problem: First near-linear time algorithm Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 24 / 26

  70. Summary • Main goal: Efficient algorithms for the Massive Parallel Computation Model Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 25 / 26

  71. Summary • Main goal: Efficient algorithms for the Massive Parallel Computation Model • Important efficiency measure: number of rounds When can it be made O ( 1 ) with low memory? Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 25 / 26

  72. Summary • Main goal: Efficient algorithms for the Massive Parallel Computation Model • Important efficiency measure: number of rounds When can it be made O ( 1 ) with low memory? • Well known obstacle: Sparse Connectivity Krzysztof Onak (IBM Research) Parallel Algorithms for Graphs on a Very Large. . . 25 / 26

Download Presentation
Download Policy: The content available on the website is offered to you 'AS IS' for your personal information and use only. It cannot be commercialized, licensed, or distributed on other websites without prior consent from the author. To download a presentation, simply click this link. If you encounter any difficulties during the download process, it's possible that the publisher has removed the file from their server.

Recommend

NC Algorithms for Computing a Perfect Matching, the Number of Perfect Matchings, and a Maximum

NC Algorithms for Computing a Perfect Matching, the Number of Perfect Matchings, and a Maximum

NC Algorithms for Computing a Perfect Matching, the Number of Perfect Matchings, and a Maximum Flow in One-Crossing-Minor-Free Graphs David Eppstein and Vijay V. Vazirani University of California, Irvine Symposium on Parallel Algorithms, June

861 views • 19 slides
Sampling Large Graphs: Algorithms and Applications Don Towsley College of Information &amp;

Sampling Large Graphs: Algorithms and Applications Don Towsley College of Information &amp;

Sampling Large Graphs: Algorithms and Applications Don Towsley College of Information &amp; Computer Science Umass - Amherst Collaborators: P.H. Wang, J.C.S. Lui, J.Z. Zhou, X. Guan Measuring, analyzing large networks - large networks can be

935 views • 52 slides
1 Graphs Graphs a a c c Graph algorithms Depth-first search, Breadth-first search b

1 Graphs Graphs a a c c Graph algorithms Depth-first search, Breadth-first search b

PhD course Bulk-Synchronous Parallel (BSP) Computing Big Data Analytics Bulk-Synchronous Parallel (BSP) Model [Valiant90] A Bridging Model between abstract PRAM model and real-world parallel computers, to support algorithm development

317 views • 5 slides
A middleware for parallel processing of large graphs Tiago Alves Macambira and Dorgival Olavo

A middleware for parallel processing of large graphs Tiago Alves Macambira and Dorgival Olavo

Outline Introduction The API Implementation Evaluation Conclusions A middleware for parallel processing of large graphs Tiago Alves Macambira and Dorgival Olavo Guedes Neto {tmacam,dorgival}@dcc.ufmg.br National Institute for Web Research

735 views • 34 slides
Graphs Graphs Simple graphs Algorithms Depth-first search Breadth-first search

Graphs Graphs Simple graphs Algorithms Depth-first search Breadth-first search

CS171 Introduction to Computer Science II Graphs Graphs Simple graphs Algorithms Depth-first search Breadth-first search shortest path Connected components Directed graphs Weighted graphs Minimum spanning tree

781 views • 53 slides
Random Graphs Omid Etesami Large graphs World Wide Web Internet Social Networks

Random Graphs Omid Etesami Large graphs World Wide Web Internet Social Networks

Topics in Algorithms and Data Science Random Graphs Omid Etesami Large graphs World Wide Web Internet Social Networks Journal Citations Economics Journals Citations Random graphs Unlike traditional graph theory, we

1.18k views • 89 slides
Parallel Algorithms for Solving Large Assignment Problems on GPU Clusters 2018 Blue Waters

Parallel Algorithms for Solving Large Assignment Problems on GPU Clusters 2018 Blue Waters

Parallel Algorithms for Solving Large Assignment Problems on GPU Clusters 2018 Blue Waters Symposium Ketan Date Rakesh Nagi (PI) Department of Industrial and Enterprise Systems Engineering University of Illinois at Urbana-Champaign June 6,

804 views • 35 slides
+ Design of Parallel Algorithms Models of Parallel Computation + Chapter Overview: Algorithms

+ Design of Parallel Algorithms Models of Parallel Computation + Chapter Overview: Algorithms

+ Design of Parallel Algorithms Models of Parallel Computation + Chapter Overview: Algorithms and Concurrency Introduction to Parallel Algorithms Tasks and Decomposition Processes and Mapping Processes Versus Processors

1.01k views • 79 slides
How to Think Algorithmically in Parallel? Or, Parallel Programming through Parallel Algorithms

How to Think Algorithmically in Parallel? Or, Parallel Programming through Parallel Algorithms

How to Think Algorithmically in Parallel? Or, Parallel Programming through Parallel Algorithms Uzi Vishkin Context Will review variant of context part in August 22 talk given at Hot Interconnects, Stanford, CA. Please relax and listen. This

1.02k views • 72 slides
Parallel Algorithms and Programming Fault tolerance for Parallel Applications Thomas Ropars

Parallel Algorithms and Programming Fault tolerance for Parallel Applications Thomas Ropars

Parallel Algorithms and Programming Fault tolerance for Parallel Applications Thomas Ropars thomas.ropars@unvi-grenoble-alpes.fr 2018 1 Agenda About failures in large scale systems The basic problem Checkpoint-based protocols Log-based

1.47k views • 119 slides
Efficient Algorithms P and NP So far, we have developed algorithms for finding shortest

Efficient Algorithms P and NP So far, we have developed algorithms for finding shortest

Efficient Algorithms P and NP So far, we have developed algorithms for finding shortest paths in graphs, CISC5835, Algorithms for Big Data minimum spanning trees in graphs, CIS, Fordham Univ. matchings in bipartite graphs,

514 views • 8 slides
PowerGraph Distributed Graph-Parallel Computation on Natural Graphs JOSHUA SEND 24/10/2017

PowerGraph Distributed Graph-Parallel Computation on Natural Graphs JOSHUA SEND 24/10/2017

PowerGraph Distributed Graph-Parallel Computation on Natural Graphs JOSHUA SEND 24/10/2017 LSDPO SESSION 3 Intuition for Graph Processing Systems Overall goal efficiently compute over large graphs of data key is distributing work

365 views • 19 slides
Parallel Numerical Algorithms Chapter 2 Parallel Thinking Section 2.3 Parallel

Parallel Numerical Algorithms Chapter 2 Parallel Thinking Section 2.3 Parallel

Efficiency Scalability Example Parallel Numerical Algorithms Chapter 2 Parallel Thinking Section 2.3 Parallel Performance Michael T. Heath and Edgar Solomonik Department of Computer Science University of Illinois at Urbana-Champaign

589 views • 38 slides
Clique Number and Chromatic Number of Graphs defined by Convex Geometries Jonathan E. Beagley

Clique Number and Chromatic Number of Graphs defined by Convex Geometries Jonathan E. Beagley

Introduction The Copoint Graph Convex Geometries Clique Number vs. Chromatic Number Copoint Graphs with Large Chromatic Number Clique Number and Chromatic Number of Graphs defined by Convex Geometries Jonathan E. Beagley Walter Morris

461 views • 34 slides
Weighted Graphs Weighted graphs: graphs for which each edge has an Minimum Spanning Trees

Weighted Graphs Weighted graphs: graphs for which each edge has an Minimum Spanning Trees

Weighted Graphs Weighted graphs: graphs for which each edge has an Minimum Spanning Trees associated weight w(u, v), a real number value CISC4080, Computer Algorithms w: E R, weight function CIS, Fordham Univ. Graph representation

468 views • 4 slides
Parallel Numerical Algorithms Chapter 2 Parallel Thinking Section 2.2 Parallel

Parallel Numerical Algorithms Chapter 2 Parallel Thinking Section 2.2 Parallel

Parallel Programming Paradigms MPI Message-Passing Interface OpenMP Portable Shared Memory Programming Parallel Numerical Algorithms Chapter 2 Parallel Thinking Section 2.2 Parallel Programming Michael T. Heath and Edgar

788 views • 45 slides
Bounded Radius Routing Perform bounded PRIM algorithm Under = 0, = 0.5, and =

Bounded Radius Routing Perform bounded PRIM algorithm Under = 0, = 0.5, and =

Bounded Radius Routing Perform bounded PRIM algorithm Under = 0, = 0.5, and = Compare radius and wirelength Radius = 12 for this net Practical Problems in VLSI Physical Design Bounded Radius Routing (1/16) BPRIM Under

542 views • 16 slides
Approximating geometrical graphs via spanners and banyans S. B. Rao &amp; W. D. Smith

Approximating geometrical graphs via spanners and banyans S. B. Rao &amp; W. D. Smith

Approximating geometrical graphs via spanners and banyans S. B. Rao &amp; W. D. Smith Proc of STOC, 1998: 540-550 Recap (Aroras Algorithm) Recap (Aroras Algorithm) Recap (Aroras Algorithm) Recap (Aroras Algorithm)

446 views • 26 slides
Algorithms R OBERT S EDGEWICK | K EVIN W AYNE 4.3 M INIMUM S PANNING T REES introduction

Algorithms R OBERT S EDGEWICK | K EVIN W AYNE 4.3 M INIMUM S PANNING T REES introduction

Algorithms R OBERT S EDGEWICK | K EVIN W AYNE 4.3 M INIMUM S PANNING T REES introduction greedy algorithm edge-weighted graph API Algorithms Kruskal's algorithm F O U R T H E D I T I O N Prim's algorithm R OBERT S EDGEWICK |

671 views • 65 slides
Deterministic MST Sparsification in the Congested Clique Janne H. Korhonen University of

Deterministic MST Sparsification in the Congested Clique Janne H. Korhonen University of

Brief Announcement: Deterministic MST Sparsification in the Congested Clique Janne H. Korhonen University of Reykjavk Introduction 1: Congested Clique Model specialisation of CONGEST communication graph = clique on n nodes input

289 views • 17 slides
CS 1501 www.cs.pitt.edu/~nlf4/cs1501/ Weighted Graphs Last time, we said spatial layouts of

CS 1501 www.cs.pitt.edu/~nlf4/cs1501/ Weighted Graphs Last time, we said spatial layouts of

CS 1501 www.cs.pitt.edu/~nlf4/cs1501/ Weighted Graphs Last time, we said spatial layouts of graphs were irrelevant We define graphs as sets of vertices and edges However, well certainly want to be able to reason about bandwidth,

466 views • 21 slides
Tangent estimation along 3D digital curves l Postolski 1 , 2 , Marcin Janaszewski 2 , Yukiko

Tangent estimation along 3D digital curves l Postolski 1 , 2 , Marcin Janaszewski 2 , Yukiko

Introduction Discrete Straight Segments -Maximal Segment Tangent Direction Experimental Validation Tangent estimation along 3D digital curves l Postolski 1 , 2 , Marcin Janaszewski 2 , Yukiko Kenmochi 1 , Micha Jacques-Olivier Lachaud 3 1

613 views • 26 slides
Algorithms R OBERT S EDGEWICK | K EVIN W AYNE K RUSKAL ' S A LGORITHM D EMO Algorithms F O U R T

Algorithms R OBERT S EDGEWICK | K EVIN W AYNE K RUSKAL ' S A LGORITHM D EMO Algorithms F O U R T

Algorithms R OBERT S EDGEWICK | K EVIN W AYNE K RUSKAL ' S A LGORITHM D EMO Algorithms F O U R T H E D I T I O N R OBERT S EDGEWICK | K EVIN W AYNE http://algs4.cs.princeton.edu Kruskal's algorithm demo Consider edges in ascending order of

373 views • 19 slides
MST &amp; Shortest-Path Algs.

MST &amp; Shortest-Path Algs.

MST &amp; Shortest-Path Algs.

809 views • 24 slides

More recommend