Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.

WebDBforum2011

2,446 views

Published on

Published in: Technology
  • Be the first to comment

WebDBforum2011

  1. 1. mixi2011 11 7
  2. 2. • (26) • • 2010 • • 2011 7 • • • ( )2011 11 7
  3. 3. Adgenda • ? • mixi •2011 11 7
  4. 4. Adgenda • • mixi •2011 11 7
  5. 5. • Web • 2007 Brad Fitzpatrick • 2007 F8 Mark Eliot Zuckerberg http://bradfitz.com/social-graph-problem/2011 11 7
  6. 6. • Web • 2007 Brad Fitzpatrick • 2007 F8 Mark Eliot Zuckerberg http://bradfitz.com/social-graph-problem/2011 11 7
  7. 7. 2011 11 7
  8. 8. Vertex (node : )2011 11 7
  9. 9. Vertex (node : ) Edge ( )2011 11 7
  10. 10. Vertex (node : ) Undirected graph ( ) Edge ( )2011 11 7
  11. 11. Vertex (node : )2011 11 7
  12. 12. Vertex (node : ) Edge ( )2011 11 7
  13. 13. Vertex (node : ) Directed graph ( ) Edge ( )2011 11 7
  14. 14. Undirected graph ( )2011 11 7
  15. 15. Vertex ( ) Undirected graph ( )2011 11 7
  16. 16. Vertex ( ) Undirected graph ( ) Edge ( )2011 11 7
  17. 17. Vertex ( ) Undirected graph ( ) Edge ( ) ( ) mixi2011 11 7
  18. 18. Vertex ( ) Directed graph ( ) Edge ( )2011 11 7
  19. 19. Vertex ( ) Directed graph ( ) Edge ( ) ( ) Twitter follow2011 11 7
  20. 20. Adgenda • ? • mixi • mixi • • • •2011 11 7
  21. 21. mixi ThinkSocial(2011 8 ) http://pr.mixi.co.jp/2011/08/15/mixiinfographic0815.html2011 11 7
  22. 22. 2011 11 7
  23. 23. 2011 11 7
  24. 24. 2011 11 7
  25. 25. 2011 11 7
  26. 26. 2011 11 7
  27. 27. 2011 11 7
  28. 28. something post I’m runnning, now!2011 11 7
  29. 29. something post I’m runnning, now!2011 11 7
  30. 30. something post I’m runnning, now! Feed Back2011 11 7
  31. 31. something post I’m runnning, now! Feed Back2011 11 7
  32. 32. 2011 11 7
  33. 33. 2011 11 7
  34. 34. •2011 11 7
  35. 35. • •2011 11 7
  36. 36. 2011 11 7
  37. 37. 2011 11 7
  38. 38. R • • • • http://www.r-project.org/2011 11 7
  39. 39. Gephi • • Google Summer Of Code • • http://gephi.org/2011 11 7
  40. 40. Adgenda • ? • mixi • • • • Girven-Newman • Gephi R2011 11 7
  41. 41. • • •2011 11 7
  42. 42. • • •2011 11 7
  43. 43. • • •2011 11 7
  44. 44. • • • • Q2011 11 7
  45. 45. • • •2011 11 7
  46. 46. • • • 2 1 1 4 2 1 22011 11 7
  47. 47. • • • 1 B A D C2011 11 7
  48. 48. • • • 1 B A 3 A D C2011 11 7
  49. 49. • • • 1 B A 3 (A,B,C) A D C2011 11 7
  50. 50. • • • 1 B A 3 (A,B,C) (A,B,D) A D C2011 11 7
  51. 51. • • • 1 B A 3 (A,B,C) (A,B,D) (A,C,D) A D C2011 11 7
  52. 52. • • • 1 B A 3 (A,B,C) (A,B,D) (A,C,D) 1 (A,B,C) A D C2011 11 7
  53. 53. • • • 1 B A 3 (A,B,C) (A,B,D) (A,C,D) 1 (A,B,C) A D = 1/3 = 0.33 C2011 11 7
  54. 54. • • : v : s, t : s,t ( )v2011 11 7
  55. 55. • • • ≠2011 11 7
  56. 56. • • • ≠2011 11 7
  57. 57. Q • • : :( i )/( ) :( i )/ ( )2011 11 7
  58. 58. 2011 11 7
  59. 59. • 1hop • 40 40 • : 0.25 • Girvan-Newman ThinkSocial(2011 8 )http://pr.mixi.co.jp/2011/08/15/mixiinfographic0815.html http://alpha.mixi.co.jp/blog/?p=1442011 11 7
  60. 60. Girvan-Newman 1. 2. (betweenness score) 3. betweenness score 4. betweenness score 5. 2,3,42011 11 7
  61. 61. Gephi 41 ( + 40 ) 2362011 11 7
  62. 62. 41 ( + 40 ) 2362011 11 7
  63. 63. R # igraph library(igraph) # Girven-Newman # g eb <- edge.betweenness.community(g) # Q step <- 1:40 for (i in step) { memb <- community.to.membership(g, eb$merges, steps=i-1) Q[i] <- modularity(g, memb$membership) } # Q plot(step, Q, type=b)2011 11 7
  64. 64. step Q2011 11 7
  65. 65. step Q 38 Q2011 11 7
  66. 66. step Q # 38 > community.to.membership(g, eb$merges, steps=38)$csize [1] 26 11 42011 11 7
  67. 67. 41 ( + 40 ) 236 32011 11 7
  68. 68. • • •2011 11 7
  69. 69. (2hop)2011 11 7
  70. 70. (2hop) 2598 3456 33 3462011 11 7
  71. 71. Girven-Newman , betweenness score2011 11 7
  72. 72. Girven-Newman # Girven-Newman # 1hop 2hop # 1hop > system.time(GirvenNewman(graph.1hop)) 0.10 0.00 0.09 # 2hop > system.time(GirvenNewman(graph.2hop)) 77.61 0.02 78.992011 11 7
  73. 73. Girven-Newman # Girven-Newman # 1hop 2hop # 1hop > system.time(GirvenNewman(graph.1hop)) 0.10 0.00 0.09 # 2hop > system.time(GirvenNewman(graph.2hop)) 77.61 0.02 78.992011 11 7
  74. 74. Girven-Newman # Girven-Newman # 1hop 2hop # 1hop > system.time(GirvenNewman(graph.1hop)) 0.10 0.00 0.09 # 2hop > system.time(GirvenNewman(graph.2hop)) 77.61 0.02 78.992011 11 7
  75. 75. • "Fast algorithm for detecting community structure in networks" • M.E.J.Newman • Phys. Rev. E 69, 066133(2004) • Newman • • ΔQ • •2011 11 7
  76. 76. # 1hop 2hop # 1hop > system.time(GirvenNewman(graph.1hop)) 0.10 0.00 0.09 > system.time(Newman(graph.1hop)) 0.08 0.00 0.04 # 2hop > system.time(GirvenNewman(graph.2hop)) 77.61 0.02 78.99 > system.time(Newman(graph.2hop)) 25.49 0.02 26.352011 11 7
  77. 77. # 1hop 2hop # 1hop > system.time(GirvenNewman(graph.1hop)) 0.10 0.00 0.09 > system.time(Newman(graph.1hop)) 0.08 0.00 0.04 # 2hop > system.time(GirvenNewman(graph.2hop)) 77.61 0.02 78.99 > system.time(Newman(graph.2hop)) 25.49 0.02 26.352011 11 7
  78. 78. # 1hop 2hop # 1hop > system.time(GirvenNewman(graph.1hop)) 0.10 0.00 0.09 > system.time(Newman(graph.1hop)) 0.08 0.00 0.04 # 2hop > system.time(GirvenNewman(graph.2hop)) 77.61 0.02 78.99 > system.time(Newman(graph.2hop)) 25.49 0.02 26.352011 11 7
  79. 79. # 1hop 2hop # 1hop > system.time(GirvenNewman(graph.1hop)) 0.10 0.00 0.09 > system.time(Newman(graph.1hop)) 0.08 0.00 0.04 # 2hop > system.time(GirvenNewman(graph.2hop)) 77.61 0.02 78.99 > system.time(Newman(graph.2hop)) 25.49 0.02 26.352011 11 7
  80. 80. (1) • "Finding community structure in very large networks" • Aaron Clauset, M.E.J.Newman, Cristopher Moore • Phys. Rev. E 70, 066111(2004) • Newman • ΔQ •2011 11 7
  81. 81. (2) • "Fast unfolding of communities in large networks" • Vincent D.Blondel, Jean-Loup Guilaume, Renaud Lambiotte, Etienne Lefebvre • J. Stat. Mech. (2008) P10008 • ΔQ • CPU : bi-opteron 2.2G, memory 24G • 1 2000 152min2011 11 7
  82. 82. • • mixi • • • • • Girven-Newman •2011 11 7
  83. 83. "Finding and evaluating community structure in networks" M.E.J.Newman and M.Girvan Phys. Rev. E 69, 026113(2004) "Fast algorithm for detecting community structure in networks" M.E.J.Newman Phys. Rev. E 69, 066133(2004) "Finding community structure in very large networks" Aaron Clauset, M.E.J.Newman, Cristopher Moore Phys. Rev. E 70, 066111(2004) "Fast unfolding of communities in large networks" Vincent D.Blondel, Jean-Loup Guilaume, Renaud Lambiotte, Etienne Lefebvre J. Stat. Mech. (2008) P100082011 11 7

×