作者: Yash Deshpande , Emmanuel Abbe , Andrea Montanari
DOI:
关键词: Stochastic block model 、 Variation of information 、 Mutual information 、 Mathematics 、 Statistical model 、 Interaction information 、 Discrete mathematics 、 Upper and lower bounds 、 Pointwise mutual information 、 Vertex (geometry) 、 Combinatorics
摘要: We develop an information-theoretic view of the stochastic block model, a popular statistical model for large-scale structure complex networks. A graph $G$ from such is generated by first assigning vertex labels at random finite alphabet, and then connecting vertices with edge probabilities depending on endpoints. In case symmetric two-group we establish explicit `single-letter' characterization per-vertex mutual information between graph. The expression intimately related to estimation-theoretic quantities, --in particular-- reveals phase transition critical point community detection. Below asymptotically same as if edges were independent. Correspondingly, no algorithm can estimate partition better than guessing. Conversely, above threshold, strictly smaller independent-edges upper bound. this regime there exists procedure that estimates