Analytic Theory of Power Law Graphs

Transcription

1 Analytic Theory of Power Law Graphs Jeremy Kepner This work is sponsored by the Department of Defense under Air Force Contract FA C Opinions, interpretations, conclusions, and recommendations are those of the author and are not necessarily endorsed by the United States Government. Slide-1

2 Outline Introduction ti Kronecker Graphs Graphs as Matrices Algorithm Comparison B K Graphs (B+I) K Graphs Summary Slide-2

3 Power Law Modeling of Kronecker Graphs Adjacency Matrix Vertex In Degree Distribution er of Vertices Numbe Power Law In Degree Real world data (internet, social networks, ) has connections on all scales (i.e power law) Can be modeled with Kronecker Graphs: G k = G k-1 G Where denotes the Kronecker product of two matrices Slide-3

4 Graphs as Matrices A T x A T x Graphs can be represented as a sparse matrices 3 6 Multiply by adjacency matrix step to neighbor vertices Work-efficient implementation from sparse data structures Most algorithms reduce to products on semi-rings: C = A +. x B x : associative, distributes over + + : associative, commutative Examples: +.* min.+ or.and Slide-4

5 Algorithm (Problem) Algorithm Comparison Canonical Complexity Array-Based Complexity Critical Path (for array) Bellman-Ford (SSSP) Θ(mn) ( ) Θ(mn) ( ) Θ(n) ( ) Generalized B-F (APSP) NA Θ(n 3 log n) Θ(log n) Floyd-Warshall (APSP) Θ(n 3 ) Θ(n 3 ) Θ(n) Prim (MST) Θ(m+n log n) Θ(n 2 ) Θ(n) Borůvka (MST) Θ(m log n) Θ(m log n) Θ(log 2 n) Edmonds-Karp (Max Flow) Θ(m 2 n) Θ(m 2 n) Θ(mn) Push-Relabel (Max Flow) Θ(mn 2 ) O(mn 2 )? (or Θ(n 3 )) Greedy MIS (MIS) Θ(m+n( log n) ) Θ(mn+n( 2 ) Θ(n) ( ) Luby (MIS) Θ(m+n log n) Θ(m log n) Θ(log n) Majority of selected algorithms can be represented with array-based constructs with equivalent complexity. Slide-5 (n = V and m = E.)

6 Outline Introduction ti B K Graphs (B+I) K Graphs Definitions Bipartite Graphs Degree Distribution Summary Slide-6

7 Kronecker Products and Graph Kronecker Product Let B be a N B xn B matrix Let C be a N C xn C matrix Then the Kronecker product of B and C will produce a N B N C xn B N C matrix A: Kronecker Graph (Leskovec 2005 & Chakrabati 2004) Let G be a NxN N adjacency matrix Kronecker exponent to the power k is: Slide-7

8 Types of Kronecker Graphs Explicit it Stochastic ti Instance Explicit G only 1 and 0s Stochastic G contains probabilities G 1 Instance A set of M points (edges) drawn from a stochastic G 2 Slide-8 G 3

9 Kronecker Product of a Bipartite Graph = P Equal with the right permutation = P B(5,1) B(3,1) = P B(15,1) 1) B(3,5) Fundamental result [Weischel 1962] is that the Kronecker product of two complete bipartite graphs is two complete bipartite graphs More generally Slide-9

10 Degree Distribution of Bipartite Kronecker Graphs Kronecker exponent of a bipartite graph produces many independent bipartite graphs Only k+1 different kinds of nodes in this graph, with degree distribution Slide-10

11 Explicit Degree Distribution Kronecker exponent of bipartite graph naturally produces exponential distribution log n (Nu umber of Vertices) B(n=10,1) k=5 B(n=5,1) k=10 Slide-11 log n (Vertex Degree)

12 Instance Degree Distribution 1M Edges B(4,1) 6 s Number of Vertices An instance graph drawn from a stochastic bipartite graph is just the sum of Poisson distributions taken from the explicit i bipartite i graph Slide-12

13 Outline Introduction ti B K Graphs (B+I) K Graphs Summary Bipartite + Identity Graphs Permutations and substructure Degree Distribution Iso Parametric Ratio Slide-13

14 Theory Bipartite Kronecker graphs highlight the fundamental structures in a Kronecker graph, but Are not connected (i.e. many independent bipartite graphs) Adding identity matrix creates connections on all scales Resulting explicit graph has diameter = 2 Sub-structures in the graph are given by Where indicates permutations are required to add the matrices Sub-structure can be revealed by applying permutation that groups vertices by their bipartite sub-graph Slide-14

15 Bipartite Permutation (B(4,1)+I) 4 P 4 ((B(4,1)+I) 4 ) Left: unpermuted (B+I) 4 kronecker graph Right: permuted (B+I) 4 kronecker graph Slide-15

16 Identifying Substructure (B+I) 3 1st+2nd Order Higher Order - = P 3 -(B 3 ) P 3 -(B 3 +B B I) P 3 -(B 3 +B I B) P 3 -(B 3 + I B B) Permuting specific terms shows their contributions to the graph Slide-16

17 Quantifying Substructure P (B 5 ) P 5 (B 5 +B 4 I B 0 ) P 5 (B 5 +B 3 I B 1 ) χ 0 5 χ 1 5 P 5 (B 5 +B 2 I B 2 ) P 5 (B 5 +B 1 I B 3 ) P 5 (B 5 +B 0 I B 4 ) χ 2 5 χ 3 5 χ 2 χ 3 χ 4 5 Connections between bipartite subgraphs are the Kronecker product of corresponding 2x2 matrices, e.g. B(1,1) 4 I(2) Slide-17

18 Substructure Degree Distribution B k B k + 2nd Order Number of Vertices (B+I) k Vertex Degree Only k+1 different kinds of nodes in this graph, with same degree distribution, only differing values of vertex degree (B+I) k is steeper than B k Slide-18

19 Example Result: Iso-Parametric Ratio Iso-Par rametric Ratio IsoPar(n r ) half of bipartite sub-graph IsoPar(n r n k-r ) all of bipartite sub-graph Sub-graph size = n r Iso-parametric ratios measure the surface to volume of a sub-graph Can analytically compute for a Kronecker graph: (B+I) k Shows large effect of including half or all of bipartite sub-graph Slide-19

20 Kronecker Graph Theory -Summary of Current Results- Quantity Graph: B(n,m) k Graph: (B+I) k Degree Distribution Betweenness Centrality Diameter Eigenvalues Iso-parametric Ratio half Iso-parametric Ratio all Slide-20

21 Reference Book: Graph Algorithms in the Language of Linear Algebra Editors: Kepner (MIT-LL) and Gilbert (UCSB) Contributors Bader (Ga Tech) Chakrabart (CMU) Dunlavy (Sandia) Faloutsos (CMU) Fineman (MIT-LL & MIT) Gilbert (UCSB) Kahn (MIT-LL & Brown) Kegelmeyer (Sandia) Kepner (MIT-LL) Kleinberg (Cornell) Kolda (Sandia) Leskovec (CMU) Madduri (Ga Tech) Robinson (MIT-LL & NEU), Shah (UCSB) Graph Algorithms in the Language of Linear Algebra Jeremy Kepner and John Gilbert (editors) = P Slide-21