rumour spreading
play

Rumour Spreading without the Network Alessandro Panconesi - PowerPoint PPT Presentation

Rumour Spreading without the Network Alessandro Panconesi Dipartimento di Informatica Joint work with: Pawel Brach, Alessandro Epasto, Piotr Sankowski THE STARS PEOPLE The INTERNET is an observatory of


  1. Rumour Spreading without the Network Alessandro Panconesi Dipartimento di Informatica Joint work with: Pawel Brach, Alessandro Epasto, Piotr Sankowski

  2. THE ¡STARS ¡

  3. PEOPLE ¡

  4. The ¡INTERNET ¡is ¡an ¡observatory ¡of ¡Crowds ¡

  5. Digital ¡Traces ¡

  6. The ¡Grand ¡Challenge ¡

  7. The Grand Challenge What can we reconstruct the original diffusion process from the huge, and yet scanty, digital traces?

  8. Rumour spreading, a case study

  9. Gossip: a very simple model

  10. Gossiping

  11. Gossiping

  12. Gossiping

  13. Gossiping

  14. Gossiping

  15. Gossiping

  16. Gossiping

  17. Gossiping Variants PUSH Node with information sends to a random neighbour

  18. Gossiping Variants PUSH Node with information sends to a random neighbour PULL Node without information asks a random neighbour

  19. Gossiping Variants PUSH-PULL PUSH Node with information sends to a random neighbour PULL Node without information asks a random neighbour

  20. The problem that we want to solve RUMOUR SPREADING WITHOUT THE NETWORK

  21. Beyond ¡the ¡asymptoBc ¡tradiBon ¡ Can we predict the number of informed nodes at time t on the basis of the degree distribution alone?

  22. Beyond ¡the ¡asymptoBc ¡tradiBon ¡ Can we predict the average number of informed nodes at time t on the basis of the degree distribution alone?

  23. Beyond ¡the ¡asymptoBc ¡tradiBon ¡ Can we predict the average number of informed nodes at time t on the basis of the degree distribution alone for real social networks ?

  24. The Master Plan

  25. The master plan • Develop in a rigorous way a space- efficient simulator for a model • Test it with real networks

  26. THE MODEL

  27. Configuration Model D = ( )

  28. Configuration Model D = ( )

  29. Configuration Model D = ( )

  30. Configuration Model D = ( )

  31. Configuration Model D = ( )

  32. Configuration Model D = ( )

  33. Configuration Model D = ( ) Is this a good model for social networks?

  34. Configuration Model D = ( ) Is this a good model for social networks? No, but this is good!

  35. Problem ¡restatement ¡ Can we predict the average number of informed nodes at time t on the basis of the degree distribution alone for the configuration model ?

  36. Problem ¡restatement ¡ Can we predict the average number of informed nodes at time t on the basis of the degree distribution alone for the configuration model ? YES, OF COURSE!

  37. Naive Simulator • On input D = (d 1 ,d 2 ,…,d n ), pick a random graph G(D) from the configuration model • Pick a random source and simulate rumour spreading • Compute averages • Repeat

  38. THE SPACE-EFFICIENT SIMULATOR

  39. The Efficient Simulator D = ( )

  40. The Efficient Simulator D = ( )

  41. The Efficient Simulator D = ( )

  42. The Efficient Simulator D = ( )

  43. The Efficient Simulator D = ( )

  44. The Efficient Simulator D = ( )

  45. The Efficient Simulator D = ( )

  46. The Efficient Simulator D = ( )

  47. The Efficient Simulator D = ( )

  48. The Efficient Simulator D = ( )

  49. The Efficient Simulator D = ( ) This is space-efficient because we do not need to keep the stubs, only their number

  50. The Efficient Simulator D = ( ) For undirected networks further optimization is possible. The resulting savings are spectacular

  51. Dealing with aggregates Rank(u) = #unused stubs of node u M[i,j] = #nodes of degree j and rank j DxD matrix

  52. Dealing with aggregates

  53. Theorem • The Efficient Simulator is a correct implementation of the Naïve Simulator-- they compute the same averages

  54. A Picture is Worth a Thousand Words

  55. EXPERIMENTS WITH REAL NETWORKS

  56. Experiments ¡with ¡real ¡networks ¡ ? Input: ¡the ¡degree ¡ Efficient ¡simulator ¡for ¡ distribuBon ¡of ¡a ¡real ¡ the ¡configuraBon ¡model ¡ network ¡

  57. The Good.. Epinions

  58. The Good..

  59. The Bad..

  60. and the Ugly

  61. Different behaviours • Friendship and trust networks: Epinions, Facebook, LiveJournal, RenRen, and Slashdot • Collaboration and Email networks: AstroPh, CondMatt, DBLP and WikiTalk; EuAll and Enron • Non-social newtorks: Web, Amazon

  62. Courtesy of Silvio Lattanzi MEASURING RANDOMNESS

  63. Sudden drops

  64. To summarize • We developed a space-efficient predictor for the configuration model • Surprisingly, this works quite well for real social networks too

  65. Future work • Look for more efficient predictors, eg systems of differential equations • Go beyond averages • Extend to other diffusion processes?

  66. THANKS

Recommend


More recommend