Step 1: Characteristics of time intervals C1: To find the dense subgraph, we only need to consider the time intervals j such that the cohesive density curve has a local maximum at certain points between i and j under ECP local maximun must be denser l 、17“19T=20 timestamps top-k for better local minimum cohesive density curve, y(x)=density(Gx C2: All dense subgraph have a non-negative cohesive density. C3: G, j with a higher positive cohesive density has a higher probability of containing a dense subgraph 12Step 1: Characteristics of time intervals cohesive density curve, y(x) = cdensity(Gx ) 12 C2: All dense subgraph have a non-negative cohesive density. C3: G[i, j] with a higher positive cohesive density has a higher probability of containing a dense subgraph. C1: To find the dense subgraph, we only need to consider the time intervals [i, j] such that the cohesive density curve has a local maximum at certain points between i and j under ECP. must be denser! top-k for better