Figure 2From: A scalable method for discovering significant subnetworksA sample collection of input graphs and the resulting template graph. (a) A collection D of m = 4 graphs. Each graph has the same node set V = {A, B, C, D} but different edge set E i . (b) An edge-weighted graph summarized from D. Weight for an edge e is fr ( e , D ) . For example, the edge (A, C) appears in two graphs in D, so, fr ( ( A , C ) , D ) =2/4=0.5. (c) The template graph for D. Weight for an edge e is Ï• ( e ) =-log ( f r ( e , D ) ) . For example, Ï• ( ( A , C ) ) =-log ( f r ( ( A , C ) , D ) ) = Â -Â log ( 0 . 5 ) =0.69.Back to article page