1. CMU SCS Mining Billion-Node Graphs - Patterns and Algorithms Christos Faloutsos CMU

2. CMU SCS Thank you! Vipin Kumar Arindam Banerjee Jennifer Olk UMN CRAY, 2012C. Faloutsos (CMU) 2

3. CMU SCS C. Faloutsos (CMU) 3 Resource Open source system for mining huge graphs: PEGASUS project (PEta GrAph mining System) www.cs.cmu.edu/~pegasus code and papers UMN CRAY, 2012

4. CMU SCS C. Faloutsos (CMU) 4 Roadmap Introduction – Motivation Problem#1: Patterns in graphs Problem#2: Tools Problem#3: Scalability Conclusions UMN CRAY, 2012

5. CMU SCS C. Faloutsos (CMU) 5 Graphs - why should we care? Internet Map [lumeta.com] Food Web [Martinez ’91] >$10B revenue >0.5B users UMN CRAY, 2012

6. CMU SCS C. Faloutsos (CMU) 6 Graphs - why should we care? IR: bi-partite graphs (doc-terms) web: hyper-text graph... and more: D1D1 DNDN T1T1 TMTM... UMN CRAY, 2012

7. CMU SCS C. Faloutsos (CMU) 7 Graphs - why should we care? ‘viral’ marketing web-log (‘blog’) news propagation computer network security: email/IP traffic and anomaly detection.... Subject-verb-object -> graph Many-to-many db relationship -> graph UMN CRAY, 2012

8. CMU SCS C. Faloutsos (CMU) 8 Outline Introduction – Motivation Problem#1: Patterns in graphs –Static graphs –Weighted graphs –Time evolving graphs Problem#2: Tools Problem#3: Scalability Conclusions UMN CRAY, 2012

9. CMU SCS C. Faloutsos (CMU) 9 Problem #1 - network and graph mining What does the Internet look like? What does FaceBook look like? What is ‘normal’/‘abnormal’? which patterns/laws hold? UMN CRAY, 2012

10. CMU SCS C. Faloutsos (CMU) 10 Problem #1 - network and graph mining What does the Internet look like? What does FaceBook look like? What is ‘normal’/‘abnormal’? which patterns/laws hold? –To spot anomalies (rarities), we have to discover patterns UMN CRAY, 2012

11. CMU SCS C. Faloutsos (CMU) 11 Problem #1 - network and graph mining What does the Internet look like? What does FaceBook look like? What is ‘normal’/‘abnormal’? which patterns/laws hold? –To spot anomalies (rarities), we have to discover patterns –Large datasets reveal patterns/anomalies that may be invisible otherwise… UMN CRAY, 2012

12. CMU SCS C. Faloutsos (CMU) 12 Graph mining Are real graphs random? UMN CRAY, 2012

13. CMU SCS C. Faloutsos (CMU) 13 Laws and patterns Are real graphs random? A: NO!! –Diameter –in- and out- degree distributions –other (surprising) patterns So, let’s look at the data UMN CRAY, 2012

14. CMU SCS C. Faloutsos (CMU) 14 Solution# S.1 Power law in the degree distribution [SIGCOMM99] log(rank) log(degree) internet domains att.com ibm.com UMN CRAY, 2012

15. CMU SCS C. Faloutsos (CMU) 15 Solution# S.1 Power law in the degree distribution [SIGCOMM99] log(rank) log(degree) -0.82 internet domains att.com ibm.com UMN CRAY, 2012

16. CMU SCS C. Faloutsos (CMU) 16 Solution# S.2: Eigen Exponent E A2: power law in the eigenvalues of the adjacency matrix E = -0.48 Exponent = slope Eigenvalue Rank of decreasing eigenvalue May 2001 UMN CRAY, 2012

17. CMU SCS C. Faloutsos (CMU) 17 Solution# S.2: Eigen Exponent E [Mihail, Papadimitriou ’02]: slope is ½ of rank exponent E = -0.48 Exponent = slope Eigenvalue Rank of decreasing eigenvalue May 2001 UMN CRAY, 2012

18. CMU SCS C. Faloutsos (CMU) 18 But: How about graphs from other domains? UMN CRAY, 2012

19. CMU SCS C. Faloutsos (CMU) 19 More power laws: web hit counts [w/ A. Montgomery] Web Site Traffic in-degree (log scale) Count (log scale) Zipf users sites ``ebay’’ UMN CRAY, 2012

20. CMU SCS C. Faloutsos (CMU) 20 epinions.com who-trusts-whom [Richardson + Domingos, KDD 2001] (out) degree count trusts-2000-people user UMN CRAY, 2012

21. CMU SCS And numerous more # of sexual contacts Income [Pareto] –’80-20 distribution’ Duration of downloads [Bestavros+] Duration of UNIX jobs (‘mice and elephants’) Size of files of a user … ‘Black swans’ UMN CRAY, 2012C. Faloutsos (CMU) 21

22. CMU SCS C. Faloutsos (CMU) 22 Roadmap Introduction – Motivation Problem#1: Patterns in graphs –Static graphs degree, diameter, eigen, triangles cliques –Weighted graphs –Time evolving graphs Problem#2: Tools UMN CRAY, 2012

23. CMU SCS C. Faloutsos (CMU) 23 Solution# S.3: Triangle ‘Laws’ Real social networks have a lot of triangles UMN CRAY, 2012

24. CMU SCS C. Faloutsos (CMU) 24 Solution# S.3: Triangle ‘Laws’ Real social networks have a lot of triangles –Friends of friends are friends Any patterns? UMN CRAY, 2012

25. CMU SCS C. Faloutsos (CMU) 25 Triangle Law: #S.3 [Tsourakakis ICDM 2008] ASNHEP-TH Epinions X-axis: # of participating triangles Y: count (~ pdf) UMN CRAY, 2012

26. CMU SCS C. Faloutsos (CMU) 26 Triangle Law: #S.3 [Tsourakakis ICDM 2008] ASNHEP-TH Epinions UMN CRAY, 2012 X-axis: # of participating triangles Y: count (~ pdf)

27. CMU SCS C. Faloutsos (CMU) 27 Triangle Law: #S.4 [Tsourakakis ICDM 2008] SNReuters Epinions X-axis: degree Y-axis: mean # triangles n friends -> ~n 1.6 triangles UMN CRAY, 2012

28. CMU SCS C. Faloutsos (CMU) 28 Triangle Law: Computations [Tsourakakis ICDM 2008] But: triangles are expensive to compute (3-way join; several approx. algos) Q: Can we do that quickly? details UMN CRAY, 2012

29. CMU SCS C. Faloutsos (CMU) 29 Triangle Law: Computations [Tsourakakis ICDM 2008] But: triangles are expensive to compute (3-way join; several approx. algos) Q: Can we do that quickly? A: Yes! #triangles = 1/6 Sum ( i 3 ) (and, because of skewness (S2), we only need the top few eigenvalues! details UMN CRAY, 2012

30. CMU SCS C. Faloutsos (CMU) 30 Triangle Law: Computations [Tsourakakis ICDM 2008] 1000x+ speed-up, >90% accuracy details UMN CRAY, 2012

31. CMU SCS Triangle counting for large graphs? Anomalous nodes in Twitter(~ 3 billion edges) [U Kang, Brendan Meeder, +, PAKDD’11] 31 UMN CRAY, 2012 31 C. Faloutsos (CMU)

32. CMU SCS Triangle counting for large graphs? Anomalous nodes in Twitter(~ 3 billion edges) [U Kang, Brendan Meeder, +, PAKDD’11] 32 UMN CRAY, 2012 32 C. Faloutsos (CMU)

33. CMU SCS Triangle counting for large graphs? Anomalous nodes in Twitter(~ 3 billion edges) [U Kang, Brendan Meeder, +, PAKDD’11] 33 UMN CRAY, 2012 33 C. Faloutsos (CMU)

34. CMU SCS Triangle counting for large graphs? Anomalous nodes in Twitter(~ 3 billion edges) [U Kang, Brendan Meeder, +, PAKDD’11] 34 UMN CRAY, 2012 34 C. Faloutsos (CMU)

35. CMU SCS EigenSpokes B. Aditya Prakash, Mukund Seshadri, Ashwin Sridharan, Sridhar Machiraju and Christos Faloutsos: EigenSpokes: Surprising Patterns and Scalable Community Chipping in Large Graphs, PAKDD 2010, Hyderabad, India, 21-24 June 2010. C. Faloutsos (CMU) 35 UMN CRAY, 2012

36. CMU SCS EigenSpokes Eigenvectors of adjacency matrix  equivalent to singular vectors (symmetric, undirected graph) 36 C. Faloutsos (CMU)UMN CRAY, 2012

37. CMU SCS EigenSpokes Eigenvectors of adjacency matrix  equivalent to singular vectors (symmetric, undirected graph) 37 C. Faloutsos (CMU)UMN CRAY, 2012 N N details

38. CMU SCS EigenSpokes Eigenvectors of adjacency matrix  equivalent to singular vectors (symmetric, undirected graph) 38 C. Faloutsos (CMU)UMN CRAY, 2012 N N details

39. CMU SCS EigenSpokes Eigenvectors of adjacency matrix  equivalent to singular vectors (symmetric, undirected graph) 39 C. Faloutsos (CMU)UMN CRAY, 2012 N N details

40. CMU SCS EigenSpokes Eigenvectors of adjacency matrix  equivalent to singular vectors (symmetric, undirected graph) 40 C. Faloutsos (CMU)UMN CRAY, 2012 N N details

41. CMU SCS EigenSpokes EE plot: Scatter plot of scores of u1 vs u2 One would expect –Many points @ origin –A few scattered ~randomly C. Faloutsos (CMU) 41 u1 u2 UMN CRAY, 2012 1 st Principal component 2 nd Principal component

42. CMU SCS EigenSpokes EE plot: Scatter plot of scores of u1 vs u2 One would expect –Many points @ origin –A few scattered ~randomly C. Faloutsos (CMU) 42 u1 u2 90 o UMN CRAY, 2012

43. CMU SCS EigenSpokes - pervasiveness Present in mobile social graph  across time and space Patent citation graph 43 C. Faloutsos (CMU)UMN CRAY, 2012

44. CMU SCS EigenSpokes - explanation Near-cliques, or near- bipartite-cores, loosely connected 44 C. Faloutsos (CMU)UMN CRAY, 2012

45. CMU SCS EigenSpokes - explanation Near-cliques, or near- bipartite-cores, loosely connected 45 C. Faloutsos (CMU)UMN CRAY, 2012

46. CMU SCS EigenSpokes - explanation Near-cliques, or near- bipartite-cores, loosely connected 46 C. Faloutsos (CMU)UMN CRAY, 2012

47. CMU SCS EigenSpokes - explanation Near-cliques, or near- bipartite-cores, loosely connected So what?  Extract nodes with high scores  high connectivity  Good “communities” spy plot of top 20 nodes 47 C. Faloutsos (CMU)UMN CRAY, 2012

48. CMU SCS Bipartite Communities! magnified bipartite community patents from same inventor(s) `cut-and-paste’ bibliography! 48 C. Faloutsos (CMU)UMN CRAY, 2012

49. CMU SCS (maybe, botnets?) Victim IPs? Botnet members? 49 C. Faloutsos (CMU)III, Aug'12

50. CMU SCS C. Faloutsos (CMU) 50 Roadmap Introduction – Motivation Problem#1: Patterns in graphs –Static graphs degree, diameter, eigen, triangles cliques –Weighted graphs –Time evolving graphs Problem#2: Tools UMN CRAY, 2012

51. CMU SCS C. Faloutsos (CMU) 51 Observations on weighted graphs? A: yes - even more ‘laws’! M. McGlohon, L. Akoglu, and C. Faloutsos Weighted Graphs and Disconnected Components: Patterns and a Generator. SIG-KDD 2008 UMN CRAY, 2012

52. CMU SCS C. Faloutsos (CMU) 52 Observation W.1: Fortification Q: How do the weights of nodes relate to degree? UMN CRAY, 2012

53. CMU SCS C. Faloutsos (CMU) 53 Observation W.1: Fortification More donors, more $ ? $10 $5 UMN CRAY, 2012 ‘Reagan’ ‘Clinton’ $7

54. CMU SCS Edges (# donors) In-weights ($) C. Faloutsos (CMU) 54 Observation W.1: fortification: Snapshot Power Law Weight: super-linear on in-degree exponent ‘iw’: 1.01 < iw < 1.26 Orgs-Candidates e.g. John Kerry, $10M received, from 1K donors More donors, even more $ $10 $5 UMN CRAY, 2012

55. CMU SCS C. Faloutsos (CMU) 55 Roadmap Introduction – Motivation Problem#1: Patterns in graphs –Static graphs –Weighted graphs –Time evolving graphs Problem#2: Tools … UMN CRAY, 2012

56. CMU SCS C. Faloutsos (CMU) 56 Problem: Time evolution with Jure Leskovec (CMU -> Stanford) and Jon Kleinberg (Cornell – sabb. @ CMU) UMN CRAY, 2012

57. CMU SCS C. Faloutsos (CMU) 57 T.1 Evolution of the Diameter Prior work on Power Law graphs hints at slowly growing diameter: –diameter ~ O(log N) –diameter ~ O(log log N) What is happening in real data? UMN CRAY, 2012

58. CMU SCS C. Faloutsos (CMU) 58 T.1 Evolution of the Diameter Prior work on Power Law graphs hints at slowly growing diameter: –diameter ~ O(log N) –diameter ~ O(log log N) What is happening in real data? Diameter shrinks over time UMN CRAY, 2012

59. CMU SCS C. Faloutsos (CMU) 59 T.1 Diameter – “Patents” Patent citation network 25 years of data @1999 –2.9 M nodes –16.5 M edges time [years] diameter UMN CRAY, 2012

60. CMU SCS C. Faloutsos (CMU) 60 T.2 Temporal Evolution of the Graphs N(t) … nodes at time t E(t) … edges at time t Suppose that N(t+1) = 2 * N(t) Q: what is your guess for E(t+1) =? 2 * E(t) UMN CRAY, 2012

61. CMU SCS C. Faloutsos (CMU) 61 T.2 Temporal Evolution of the Graphs N(t) … nodes at time t E(t) … edges at time t Suppose that N(t+1) = 2 * N(t) Q: what is your guess for E(t+1) =? 2 * E(t) A: over-doubled! –But obeying the ``Densification Power Law’’ UMN CRAY, 2012

62. CMU SCS C. Faloutsos (CMU) 62 T.2 Densification – Patent Citations Citations among patents granted @1999 –2.9 M nodes –16.5 M edges Each year is a datapoint N(t) E(t) 1.66 UMN CRAY, 2012

63. CMU SCS C. Faloutsos (CMU) 63 Roadmap Introduction – Motivation Problem#1: Patterns in graphs –Static graphs –Weighted graphs –Time evolving graphs Problem#2: Tools … UMN CRAY, 2012

64. CMU SCS C. Faloutsos (CMU) 64 T.3 : popularity over time Post popularity drops-off – exponentially? lag: days after post # in links 1 2 3 @t @t + lag UMN CRAY, 2012

65. CMU SCS C. Faloutsos (CMU) 65 T.3 : popularity over time Post popularity drops-off – exponentially? POWER LAW! Exponent? # in links (log) days after post (log) UMN CRAY, 2012

66. CMU SCS C. Faloutsos (CMU) 66 T.3 : popularity over time Post popularity drops-off – exponentially? POWER LAW! Exponent? -1.6 close to -1.5: Barabasi’s stack model and like the zero-crossings of a random walk # in links (log) -1.6 days after post (log) UMN CRAY, 2012

67. CMU SCS UMN CRAY, 2012C. Faloutsos (CMU) 67 -1.5 slope J. G. Oliveira & A.-L. Barabási Human Dynamics: The Correspondence Patterns of Darwin and Einstein. Nature 437, 1251 (2005). [PDF]PDF Response time (log) Prob(RT > x) (log)

68. CMU SCS T.4: duration of phonecalls Surprising Patterns for the Call Duration Distribution of Mobile Phone Users Pedro O. S. Vaz de Melo, Leman Akoglu, Christos Faloutsos, Antonio A. F. Loureiro PKDD 2010 UMN CRAY, 2012C. Faloutsos (CMU) 68

69. CMU SCS Probably, power law (?) UMN CRAY, 2012C. Faloutsos (CMU) 69 ??

70. CMU SCS No Power Law! UMN CRAY, 2012C. Faloutsos (CMU) 70

71. CMU SCS ‘TLaC: Lazy Contractor’ The longer a task (phonecall) has taken, The even longer it will take UMN CRAY, 2012C. Faloutsos (CMU) 71 Odds ratio= Casualties(<x): Survivors(>=x) == power law

72. CMU SCS 72 Data Description Data from a private mobile operator of a large city 4 months of data 3.1 million users more than 1 billion phone records Over 96% of ‘talkative’ users obeyed a TLAC distribution (‘talkative’: >30 calls) UMN CRAY, 2012C. Faloutsos (CMU)

73. CMU SCS Outliers: UMN CRAY, 2012C. Faloutsos (CMU) 73

74. CMU SCS C. Faloutsos (CMU) 74 Roadmap Introduction – Motivation Problem#1: Patterns in graphs Problem#2: Tools –Belief Propagation –Tensors –Spike analysis Problem#3: Scalability Conclusions UMN CRAY, 2012

75. CMU SCS UMN CRAY, 2012C. Faloutsos (CMU) 75 E-bay Fraud detection w/ Polo Chau & Shashank Pandit, CMU [www’07]

76. CMU SCS UMN CRAY, 2012C. Faloutsos (CMU) 76 E-bay Fraud detection

77. CMU SCS UMN CRAY, 2012C. Faloutsos (CMU) 77 E-bay Fraud detection

78. CMU SCS UMN CRAY, 2012C. Faloutsos (CMU) 78 E-bay Fraud detection - NetProbe

79. CMU SCS UMN CRAY, 2012C. Faloutsos (CMU) 79 E-bay Fraud detection - NetProbe FAH F99% A H49% Compatibility matrix heterophily

80. CMU SCS Popular press And less desirable attention: E-mail from ‘Belgium police’ (‘copy of your code?’) UMN CRAY, 2012C. Faloutsos (CMU) 80

81. CMU SCS C. Faloutsos (CMU) 81 Roadmap Introduction – Motivation Problem#1: Patterns in graphs Problem#2: Tools –Belief Propagation –Tensors –Spike analysis Problem#3: Scalability Conclusions UMN CRAY, 2012

82. CMU SCS GigaTensor: Scaling Tensor Analysis Up By 100 Times – Algorithms and Discoveries U Kang Christos Faloutsos KDD’12 Evangelos Papalexakis Abhay Harpale UMN CRAY, 2012 82 C. Faloutsos (CMU)

83. CMU SCS Background: Tensor Tensors (=multi-dimensional arrays) are everywhere –Hyperlinks &anchor text [Kolda+,05] URL 1 URL 2 Anchor Text Java C++ C# 1 1 1 1 1 1 1 UMN CRAY, 2012 83 C. Faloutsos (CMU)

84. CMU SCS Background: Tensor Tensors (=multi-dimensional arrays) are everywhere –Sensor stream (time, location, type) –Predicates (subject, verb, object) in knowledge base “Barrack Obama is the president of U.S.” “Eric Clapton plays guitar” (26M) (48M) NELL (Never Ending Language Learner) data Nonzeros =144M UMN CRAY, 2012 84 C. Faloutsos (CMU)

85. CMU SCS all I learned on tensors: from UMN CRAY, 2012C. Faloutsos (CMU) 85 Nikos Sidiropoulos UMN Tamara Kolda, Sandia Labs (tensor toolbox)

86. CMU SCS Problem Definition How to decompose a billion-scale tensor? –Corresponds to SVD in 2D case UMN CRAY, 2012 86 C. Faloutsos (CMU)

87. CMU SCS Problem Definition  Q1: Dominant concepts/topics?  Q2: Find synonyms to a given noun phrase?  (and how to scale up: |data| > RAM) (26M) (48M) NELL (Never Ending Language Learner) data Nonzeros =144M UMN CRAY, 2012 87 C. Faloutsos (CMU)

88. CMU SCS Experiments GigaTensor solves 100x larger problem Number of nonzero = I / 50 (J) (I) (K) GigaTensor Tensor Toolbox Out of Memory 100x UMN CRAY, 2012 88 C. Faloutsos (CMU)

89. CMU SCS A1: Concept Discovery Concept Discovery in Knowledge Base UMN CRAY, 2012 89 C. Faloutsos (CMU)

90. CMU SCS A1: Concept Discovery UMN CRAY, 2012 90 C. Faloutsos (CMU)

91. CMU SCS A2: Synonym Discovery UMN CRAY, 2012 91 C. Faloutsos (CMU)

92. CMU SCS Next steps: tensors with side-info UMN CRAY, 2012C. Faloutsos (CMU) 92 George Karypis UMN Nikos Sidiropoulos UMN Tom Mitchell CMU

93. CMU SCS C. Faloutsos (CMU) 93 Roadmap Introduction – Motivation Problem#1: Patterns in graphs Problem#2: Tools –Belief propagation –Tensors –Spike analysis Problem#3: Scalability -PEGASUS Conclusions UMN CRAY, 2012

94. CMU SCS Meme (# of mentions in blogs) –short phrases Sourced from U.S. politics in 2008 94 “you can put lipstick on a pig” “yes we can” Rise and fall patterns in social media C. Faloutsos (CMU)UMN CRAY, 2012

95. CMU SCS Rise and fall patterns in social media 95 Can we find a unifying model, which includes these patterns? four classes on YouTube [Crane et al. ’08] six classes on Meme [Yang et al. ’11] C. Faloutsos (CMU)UMN CRAY, 2012

96. CMU SCS Rise and fall patterns in social media 96 Answer: YES! We can represent all patterns by single model C. Faloutsos (CMU)UMN CRAY, 2012 In Matsubara+ SIGKDD 2012

97. CMU SCS 97 Main idea - SpikeM -1. Un-informed bloggers (uninformed about rumor) -2. External shock at time n b (e.g, breaking news) -3. Infection (word-of-mouth) Time n=0Time n=n b β C. Faloutsos (CMU)UMN CRAY, 2012 Infectiveness of a blog-post at age n: -Strength of infection (quality of news) -Decay function Time n=n b +1

98. CMU SCS 98 -1. Un-informed bloggers (uninformed about rumor) -2. External shock at time n b (e.g, breaking news) -3. Infection (word-of-mouth) Time n=0Time n=n b β C. Faloutsos (CMU)UMN CRAY, 2012 Infectiveness of a blog-post at age n: -Strength of infection (quality of news) -Decay function Time n=n b +1 Main idea - SpikeM

99. CMU SCS SpikeM - with periodicity Full equation of SpikeM 99 Periodicity noon Peak 3am Dip Time n Bloggers change their activity over time (e.g., daily, weekly, yearly) Bloggers change their activity over time (e.g., daily, weekly, yearly) activity C. Faloutsos (CMU)UMN CRAY, 2012

100. CMU SCS Details Analysis – exponential rise and power-raw fall 100 Lin-log Log-log Rise-part SI -> exponential SpikeM -> exponential Rise-part SI -> exponential SpikeM -> exponential C. Faloutsos (CMU)UMN CRAY, 2012

101. CMU SCS Details Analysis – exponential rise and power-raw fall 101 Lin-log Log-log Fall-part SI -> exponential SpikeM -> power law Fall-part SI -> exponential SpikeM -> power law C. Faloutsos (CMU)UMN CRAY, 2012

102. CMU SCS Tail-part forecasts 102 SpikeM can capture tail part C. Faloutsos (CMU)UMN CRAY, 2012

103. CMU SCS “What-if” forecasting 103 e.g., given (1) first spike, (2) release date of two sequel movies (3) access volume before the release date ? ? ? ? (1) First spike (2) Release date (3) Two weeks before release C. Faloutsos (CMU)UMN CRAY, 2012

104. CMU SCS “What-if” forecasting 104 –SpikeM can forecast not only tail-part, but also rise-part ! SpikeM can forecast shape of upcoming spikes (1) First spike (2) Release date (3) Two weeks before release C. Faloutsos (CMU)UMN CRAY, 2012

105. CMU SCS C. Faloutsos (CMU) 105 Roadmap Introduction – Motivation Problem#1: Patterns in graphs Problem#2: Tools –Belief propagation –Spike analysis –Tensors Problem#3: Scalability -PEGASUS Conclusions UMN CRAY, 2012

106. CMU SCS UMN CRAY, 2012C. Faloutsos (CMU) 106 Scalability Google: > 450,000 processors in clusters of ~2000 processors each [ Barroso, Dean, Hölzle, “Web Search for a Planet: The Google Cluster Architecture” IEEE Micro 2003 ] Yahoo: 5Pb of data [Fayyad, KDD’07] Problem: machine failures, on a daily basis How to parallelize data mining tasks, then? A: map/reduce – hadoop (open-source clone) http://hadoop.apache.org/ http://hadoop.apache.org/

107. CMU SCS C. Faloutsos (CMU) 107 CentralizedHadoop/PEG ASUS Degree Distr. old Pagerank old Diameter/ANF old HERE Conn. Comp old HERE TrianglesdoneHERE Visualizationstarted Roadmap – Algorithms & results UMN CRAY, 2012

108. CMU SCS HADI for diameter estimation Radius Plots for Mining Tera-byte Scale Graphs U Kang, Charalampos Tsourakakis, Ana Paula Appel, Christos Faloutsos, Jure Leskovec, SDM’10 Naively: diameter needs O(N**2) space and up to O(N**3) time – prohibitive (N~1B) Our HADI: linear on E (~10B) –Near-linear scalability wrt # machines –Several optimizations -> 5x faster C. Faloutsos (CMU) 108 UMN CRAY, 2012

109. CMU SCS ???? 19+ [Barabasi+] 109 C. Faloutsos (CMU) Radius Count UMN CRAY, 2012 ~1999, ~1M nodes

110. CMU SCS YahooWeb graph (120Gb, 1.4B nodes, 6.6 B edges) Largest publicly available graph ever studied. ???? 19+ [Barabasi+] 110 C. Faloutsos (CMU) Radius Count UMN CRAY, 2012 ?? ~1999, ~1M nodes

111. CMU SCS YahooWeb graph (120Gb, 1.4B nodes, 6.6 B edges) Largest publicly available graph ever studied. ???? 19+? [Barabasi+] 111 C. Faloutsos (CMU) Radius Count UMN CRAY, 2012 14 (dir.) ~7 (undir.)

112. CMU SCS YahooWeb graph (120Gb, 1.4B nodes, 6.6 B edges) 7 degrees of separation (!) Diameter: shrunk ???? 19+? [Barabasi+] 112 C. Faloutsos (CMU) Radius Count UMN CRAY, 2012 14 (dir.) ~7 (undir.)

113. CMU SCS YahooWeb graph (120Gb, 1.4B nodes, 6.6 B edges) Q: Shape? ???? 113 C. Faloutsos (CMU) Radius Count UMN CRAY, 2012 ~7 (undir.)

114. CMU SCS 114 C. Faloutsos (CMU) YahooWeb graph (120Gb, 1.4B nodes, 6.6 B edges) effective diameter: surprisingly small. Multi-modality (?!) UMN CRAY, 2012

115. CMU SCS Radius Plot of GCC of YahooWeb. 115 C. Faloutsos (CMU)UMN CRAY, 2012

116. CMU SCS 116 C. Faloutsos (CMU) YahooWeb graph (120Gb, 1.4B nodes, 6.6 B edges) effective diameter: surprisingly small. Multi-modality: probably mixture of cores. UMN CRAY, 2012

117. CMU SCS 117 C. Faloutsos (CMU) YahooWeb graph (120Gb, 1.4B nodes, 6.6 B edges) effective diameter: surprisingly small. Multi-modality: probably mixture of cores. UMN CRAY, 2012 EN ~7 Conjecture: DE BR

118. CMU SCS 118 C. Faloutsos (CMU) YahooWeb graph (120Gb, 1.4B nodes, 6.6 B edges) effective diameter: surprisingly small. Multi-modality: probably mixture of cores. UMN CRAY, 2012 ~7 Conjecture:

119. CMU SCS Running time - Kronecker and Erdos-Renyi Graphs with billions edges. details

120. CMU SCS C. Faloutsos (CMU) 120 CentralizedHadoop/PEG ASUS Degree Distr. old Pagerank old Diameter/ANF old HERE Conn. Comp old HERE Triangles HERE Visualizationstarted Roadmap – Algorithms & results UMN CRAY, 2012

121. CMU SCS Generalized Iterated Matrix Vector Multiplication (GIMV) C. Faloutsos (CMU) 121 PEGASUS: A Peta-Scale Graph Mining System - Implementation and ObservationsSystem - Implementation and Observations. U Kang, Charalampos E. Tsourakakis, and Christos Faloutsos. (ICDM) 2009, Miami, Florida, USA. Best Application Paper (runner-up).ICDM UMN CRAY, 2012

122. CMU SCS Generalized Iterated Matrix Vector Multiplication (GIMV) C. Faloutsos (CMU) 122 PageRank proximity (RWR) Diameter Connected components (eigenvectors, Belief Prop. … ) Matrix – vector Multiplication (iterated) UMN CRAY, 2012 details

123. CMU SCS 123 Example: GIM-V At Work Connected Components – 4 observations: Size Count C. Faloutsos (CMU) UMN CRAY, 2012

124. CMU SCS 124 Example: GIM-V At Work Connected Components Size Count C. Faloutsos (CMU) UMN CRAY, 2012 1) 10K x larger than next

125. CMU SCS 125 Example: GIM-V At Work Connected Components Size Count C. Faloutsos (CMU) UMN CRAY, 2012 2) ~0.7B singleton nodes

126. CMU SCS 126 Example: GIM-V At Work Connected Components Size Count C. Faloutsos (CMU) UMN CRAY, 2012 3) SLOPE!

127. CMU SCS 127 Example: GIM-V At Work Connected Components Size Count 300-size cmpt X 500. Why? 1100-size cmpt X 65. Why? C. Faloutsos (CMU) UMN CRAY, 2012 4) Spikes!

128. CMU SCS 128 Example: GIM-V At Work Connected Components Size Count suspicious financial-advice sites (not existing now) C. Faloutsos (CMU) UMN CRAY, 2012

129. CMU SCS 129 GIM-V At Work Connected Components over Time LinkedIn: 7.5M nodes and 58M edges Stable tail slope after the gelling point C. Faloutsos (CMU)UMN CRAY, 2012

130. CMU SCS C. Faloutsos (CMU) 130 Roadmap Introduction – Motivation Problem#1: Patterns in graphs Problem#2: Tools Problem#3: Scalability Conclusions UMN CRAY, 2012

131. CMU SCS C. Faloutsos (CMU) 131 OVERALL CONCLUSIONS – low level: Several new patterns (fortification, triangle-laws, conn. components, etc) New tools: –belief propagation, gigaTensor, etc Scalability: PEGASUS / hadoop UMN CRAY, 2012

132. CMU SCS C. Faloutsos (CMU) 132 OVERALL CONCLUSIONS – high level BIG DATA: Large datasets reveal patterns/outliers that are invisible otherwise UMN CRAY, 2012

133. CMU SCS ML & Stats. Comp. Systems Theory & Algo. Biology Econ. Social Science Physics 133 Graph Analytics UMN CRAY, 2012C. Faloutsos (CMU)

134. CMU SCS C. Faloutsos (CMU) 134 References Leman Akoglu, Christos Faloutsos: RTG: A Recursive Realistic Graph Generator Using Random Typing. ECML/PKDD (1) 2009: 13-28 Deepayan Chakrabarti, Christos Faloutsos: Graph mining: Laws, generators, and algorithms. ACM Comput. Surv. 38(1): (2006) UMN CRAY, 2012

135. CMU SCS C. Faloutsos (CMU) 135 References Deepayan Chakrabarti, Yang Wang, Chenxi Wang, Jure Leskovec, Christos Faloutsos: Epidemic thresholds in real networks. ACM Trans. Inf. Syst. Secur. 10(4): (2008) Deepayan Chakrabarti, Jure Leskovec, Christos Faloutsos, Samuel Madden, Carlos Guestrin, Michalis Faloutsos: Information Survival Threshold in Sensor and P2P Networks. INFOCOM 2007: 1316-1324 UMN CRAY, 2012

136. CMU SCS C. Faloutsos (CMU) 136 References Christos Faloutsos, Tamara G. Kolda, Jimeng Sun: Mining large graphs and streams using matrix and tensor tools. Tutorial, SIGMOD Conference 2007: 1174 UMN CRAY, 2012

137. CMU SCS C. Faloutsos (CMU) 137 References T. G. Kolda and J. Sun. Scalable Tensor Decompositions for Multi-aspect Data Mining. In: ICDM 2008, pp. 363-372, December 2008. UMN CRAY, 2012

138. CMU SCS C. Faloutsos (CMU) 138 References Jure Leskovec, Jon Kleinberg and Christos Faloutsos Graphs over Time: Densification Laws, Shrinking Diameters and Possible Explanations, KDD 2005 (Best Research paper award). Jure Leskovec, Deepayan Chakrabarti, Jon M. Kleinberg, Christos Faloutsos: Realistic, Mathematically Tractable Graph Generation and Evolution, Using Kronecker Multiplication. PKDD 2005: 133-145 UMN CRAY, 2012

139. CMU SCS C. Faloutsos (CMU) 139 References Jimeng Sun, Yinglian Xie, Hui Zhang, Christos Faloutsos. Less is More: Compact Matrix Decomposition for Large Sparse Graphs, SDM, Minneapolis, Minnesota, Apr 2007. Jimeng Sun, Spiros Papadimitriou, Philip S. Yu, and Christos Faloutsos, GraphScope: Parameter- free Mining of Large Time-evolving Graphs ACM SIGKDD Conference, San Jose, CA, August 2007 UMN CRAY, 2012

140. CMU SCS References Jimeng Sun, Dacheng Tao, Christos Faloutsos: Beyond streams and graphs: dynamic tensor analysis. KDD 2006: 374- 383 UMN CRAY, 2012C. Faloutsos (CMU) 140

141. CMU SCS C. Faloutsos (CMU) 141 References Hanghang Tong, Christos Faloutsos, and Jia-Yu Pan, Fast Random Walk with Restart and Its Applications, ICDM 2006, Hong Kong. Hanghang Tong, Christos Faloutsos, Center-Piece Subgraphs: Problem Definition and Fast Solutions, KDD 2006, Philadelphia, PA UMN CRAY, 2012

142. CMU SCS C. Faloutsos (CMU) 142 References Hanghang Tong, Christos Faloutsos, Brian Gallagher, Tina Eliassi-Rad: Fast best-effort pattern matching in large attributed graphs. KDD 2007: 737-746 UMN CRAY, 2012

143. CMU SCS C. Faloutsos (CMU) 143 Project info UMN CRAY, 2012 Thanks to: NSF IIS-0705359, IIS-0534205, CTA-INARC ; Yahoo (M45), LLNL, IBM, SPRINT, Google, INTEL, HP, iLab www.cs.cmu.edu/~pegasus

144. CMU SCS C. Faloutsos (CMU) 144 Cast Akoglu, Leman Chau, Polo Kang, U McGlohon, Mary Tong, Hanghang Prakash, Aditya UMN CRAY, 2012 Koutra, Danae Beutel, Alex Papalexakis, Vagelis

145. CMU SCS C. Faloutsos (CMU) 145 OVERALL CONCLUSIONS – high level BIG DATA: Large datasets reveal patterns/outliers that are invisible otherwise UMN CRAY, 2012