复杂网络与度分割.pptVIP

Edmond ChowCenter for Applied Scientific ComputingLawrence Livermore National LaboratoryFebruary 2, 2004 Regular networks vs. Complex networks Complex networks Irregular structure of connections Short average path lengths Special degree distributions High clustering coefficient Neural networks Metabolic networks World-Wide Web Internet Electric power grid Social networks Telephone call graph Citation networks Graph and network structures can model many complex, multicomponent systems Complex networks have short average path lengths Metabolic networks Internet World-Wide Web Social networks Electric power grid Citation networks Phone-call networks Protein-Protein networks Food web Neural networks They are unlike “regular” networks Graph and network structures can model many complex, multicomponent systems Complex networks have short average path lengths Metabolic networks Internet World-Wide Web Social networks Electric power grid Citation networks Phone-call networks Protein-Protein networks Food web Neural networks They are unlike “regular” networks Outline Properties of complex networks “Six degrees of separation” problem Heuristic for greedy graph search A* search and heuristic Graph level heuristic Low average path length Power-law degree distributions Degree correlations:In social networks, vertices have a high probability of being connected to other vertices that have high degree Barabasi-Albert Scale Free Model Begin with a small clique Add vertices one at a time For each vertex, add edges, linking to another vertex with probability proportional to its degree (preferential attachment) Clustering coefficient Two neighboring vertices are likely to have many of the same neighbors Clustering coefficient: Milgram et al.’s 1969 experiment Try to send a letter to a person in Boston via forwarding the letter to a friend Senders from Omaha, Nebraska, and Boston Of the completed chains (20%), average chain length was 6.5 How are people able