DGE : Influence Minimization in Networks

The diffusion of undesirable phenomena over social, information and technological networks is a common problem in different domains. Domain experts typically intervene to advise solutions that mitigate the spread in question. However, these solutions are temporary, in that they only tackle the spread of a specific instance over the network, failing to account for the complex modalities of diffusion processes. We propose an optimization formulation for the problem of minimizing the spread of influence in a network by removing some of its edges. We show that the corresponding objective function is supermodular under the linear threshold model, allowing for a greedy approximate solution with provable guarantees. Preliminary experiments on real and synthetic network data show that our method significantly outperforms other common heuristics.