Improved Degree Bounds and Full Spectrum Power Laws in Preferential Attachment Networks

Consider a random preferential attachment model G(p) for network evolution that allows both node and edge arrivals. Starting with an arbitrary nonempty graph G0, at each time step, there are two possible events: with probability p > 0 a new node arrives and a new edge is added between the new node and an existing node, and with probability 1 - p a new edge is added between two existing nodes. In both cases, the involved existing nodes are chosen at random according to preferential attachment, i.e., with probability proportional to their degree. G(p) is known to generate power law networks, i.e., the fraction of nodes with degree k is proportional to k-β. Here β=(4-p)/(2-p) is in the range (2,3]. Denoting the number of nodes of degree k at time t by mk,t, we significantly improve some long-standing results. In particular, we show that mk,t is concentrated around its mean with a deviation of O(√t), which is independent of k. We also tightly bound the expectation Emk,t with an additive error of O(1/k), which is independent of t. These new bounds allow us to tightly estimate mk,t for a considerably larger k values than before. This, in turn, enables us to estimate other important quantities, e.g., the size of the k-rich club, namely, the set of all nodes with a degree at least k. Finally, we introduce a new generalized model, G(pt, rt, qt), which extends G(p) by allowing also time-varying probabilities for node and edge arrivals, as well as the formation of new components. We show that the extended model can produce power law networks with any exponent β in the range (1,∞). Furthermore, the concentration bounds established for mk,t in G(p) also apply in G(pt, rt, qt).

[1]  Jon M. Kleinberg,et al.  The Web as a Graph: Measurements, Models, and Methods , 1999, COCOON.

[2]  Colin Cooper,et al.  Scale‐free graphs of increasing degree , 2011, Random Struct. Algorithms.

[3]  J. Doob Regularity properties of certain families of chance variables , 1940 .

[4]  Mark E. J. Newman,et al.  Power-Law Distributions in Empirical Data , 2007, SIAM Rev..

[5]  J. Mattick,et al.  Accelerating, hyperaccelerating, and decelerating networks. , 2005, Physical review. E, Statistical, nonlinear, and soft matter physics.

[6]  Albert,et al.  Emergence of scaling in random networks , 1999, Science.

[7]  S N Dorogovtsev,et al.  Effect of the accelerating growth of communications networks on their structure. , 2000, Physical review. E, Statistical, nonlinear, and soft matter physics.

[8]  S. N. Dorogovtsev,et al.  Evolution of networks , 2001, cond-mat/0106144.

[9]  Derek de Solla Price,et al.  A general theory of bibliometric and other cumulative advantage processes , 1976, J. Am. Soc. Inf. Sci..

[10]  Christos Faloutsos,et al.  Graph evolution: Densification and shrinking diameters , 2006, TKDD.

[11]  Sergey N. Dorogovtsev,et al.  Accelerated growth of networks , 2002, ArXiv.

[12]  David Peleg,et al.  Core-Periphery in Networks: An Axiomatic Approach , 2014, ArXiv.

[13]  Eli Upfal,et al.  The Web as a graph , 2000, PODS.

[14]  H. Simon,et al.  ON A CLASS OF SKEW DISTRIBUTION FUNCTIONS , 1955 .

[15]  Fan Chung Graham,et al.  Random evolution in massive graphs , 2001, Proceedings 2001 IEEE International Conference on Cluster Computing.

[16]  Stephen P. Boyd,et al.  Convex Optimization , 2004, Algorithms and Theory of Computation Handbook.

[17]  F. Chung,et al.  Complex Graphs and Networks , 2006 .

[18]  Alan M. Frieze,et al.  A General Model of Undirected Web Graphs , 2001, ESA.

[19]  David Peleg,et al.  Core Size and Densification in Preferential Attachment Networks , 2015, ICALP.

[20]  Mark Newman,et al.  Networks: An Introduction , 2010 .