作者: Jure Leskovec , Andrew Tomkins , Deepayan Chakrabarti , Christos Faloutsos , Ana Paula Appel
DOI:
关键词:
摘要: Graphs appear in several settings, like social networks, recommendation systems, computer communication gene/protein biological among others. A deep, recurring question is “What do real graphs look like?” That is, how can we separate ones from synthetic or with masked portions? The main contribution of this paper ShatterPlots, a simple and powerful algorithm to extract patterns that help us spot fake/masked graphs. idea shatter graph, by deleting edges, force it reach critical (“Shattering”) point, study the properties at point. One most striking “30-per-cent”: Shattering all have about 30% more nodes than edges. our discriminative “NodeShatteringRatio ”, which almost perfectly extensive collection. Additional contributions are (a) careful, scalable design algorithm, requires only O(E) time, (b) experiments large collection (19 total), up hundreds thousands million (c) wealth observations patterns, show distinguish ones.