EPIC: Welfare Maximization under Economically Postulated Independent Cascade Model.

arXiv (Cornell University)(2018)

引用 0|浏览74
In economics, it is well accepted that adoption of items is governed by the utility that a user derives from their adoption. In this paper, we propose a model called EPIC that combines utility-driven item adoption with the viral network effect helping to propagate adoption of and desire for items from users to their peers. We focus on the case of mutually complementary items and model their adoption behavior via supermodular value functions. We assume price is additive and use zero mean random noise to capture the uncertainty in our knowledge of user valuations. In this setting, we study a novel problem of \emph{social welfare maximization}: given item budgets, find an optimal allocation of items to seed nodes that maximizes the sum of expected utilities derived by users when the diffusion terminates. We show the expected social welfare is monotone but neither submodular nor supermodular. Nevertheless, we show that a simple greedy allocation can ensure a $(1-1/e-\epsilon)$-approximation to the optimum. To the best of our knowledge, this is the first instance where for a non-submodular objective in the context of viral marketing, such a high approximation ratio is achieved. We provide the analysis of this result, which is highly nontrivial and along the way we give a solution to the prefix-preserving influence maximization problem, which could be of independent interest. With extensive experiments on real and synthetic datasets, we show that our algorithm significantly outperforms all the baselines.
welfare maximization
AI 理解论文
Chat Paper