Blocking Influence at Collective Level with Hard Constraints (Student Abstract)

Abstract

Influence blocking maximization (IBM) is crucial in many critical real-world problems such as rumors prevention and epidemic containment. The existing work suffers from: (1) concentrating on uniform costs at the individual level, (2) mostly utilizing greedy approaches to approximate optimization, (3) lacking a proper graph representation for influence estimates. To address these issues, this research introduces a neural network model dubbed Neural Influence Blocking (\algo) for improved approximation and enhanced influence blocking effectiveness. The code is available at https://github.com/oates9895/NIB.

Explain Influence Maximization with Sobol Indices

“…(3) Pi [50]: Considering the existence of cycles in networks, a node can be reached multiple times by a seed in different step lengths. Thus, the nodes spreading powers can be better estimated by…”

“…The time for each iteration of the node selection and proxy update remains the same while the number of iterations raises from k to ak. For example, the time complexity of the Pi IM algorithm [50] is O(k • n 3 ). The corresponding time complexity for SobolIM 's first step is O(ak • n 3 ).…”

“…To alleviate the heavy computational burden of simulations, some researchers turned to proxy-based methods in which the spreading power of the nodes is estimated by certain proxies. They started from simple heuristic measures such as degree, PageRank [30] and eigen-centrality [51] and later turned to several influence-aware or diffusion model-aware proxies [5,23,6,7,42,15,46,50] to better estimate the influence effect brought by the seed. However, due to the intricate interactions among the nodes within the seed set [22,5,26], the actual influence effect of seeds selected by those proxies are not guaranteed to match the proxies' initial intention (which we will show in the experiments).…”