Main ideas Big graph friendly Employ hidden data statistics to explore k time intervals k is typically a small constant independent of T, e.g., 10 T 141 447 1,414 14,142 T*(T+12 10 106 108 unpruned 10 105 our approach k Our data-driven approach fIDEs step 1: identify k time intervals involved with dense subgraphs employing hidden data statistics and drawing characteristics of targeted time intervals step 2: compute dense subgraphs given time intervals v building the connections with the NWM problem, and exploiting effective and efficient optimization techniques 1000x faster while remain comparable quality of dense subgraphs 10Main ideas ➢ Employ hidden data statistics to explore k time intervals • k is typically a small constant independent of T, e.g., 10 10 T 141 447 1,414 ··· 14,142 T*(T+1)/2 104 105 106 ··· 108 # unpruned 102 103 104 ··· 106 our approach k k k … k ➢ Our data-driven approach FIDES • step 1: identify k time intervals involved with dense subgraphs ✓ employing hidden data statistics and drawing characteristics of targeted time intervals • step 2: compute dense subgraphs given time intervals ✓ building the connections with the NWM problem, and exploiting effective and efficient optimization techniques Big graph friendly 1000x faster while remain comparable quality of dense subgraphs