Scalable Lattice Influence Maximization

Influence maximization is the task of finding <inline-formula> <tex-math notation="LaTeX">$k$ </tex-math></inline-formula> seed nodes in a social network such that the expected number of activated nodes in the network (under certain influence propagation model), referred to as the influence spread, is maximized. Lattice influence maximization (LIM) generalizes influence maximization such that, instead of selecting <inline-formula> <tex-math notation="LaTeX">$k$ </tex-math></inline-formula> seed nodes, one selects a vector <inline-formula> <tex-math notation="LaTeX">$\mathbf {x}=(x_{1}, \ldots, x_{d})$ </tex-math></inline-formula> from a discrete space <inline-formula> <tex-math notation="LaTeX">$\mathcal {X}$ </tex-math></inline-formula> called a lattice, where <inline-formula> <tex-math notation="LaTeX">$x_{j}$ </tex-math></inline-formula> corresponds to the <inline-formula> <tex-math notation="LaTeX">$j$ </tex-math></inline-formula>th marketing strategy and <inline-formula> <tex-math notation="LaTeX">$\mathbf {x}$ </tex-math></inline-formula> represents a marketing strategy mix. Each strategy mix <inline-formula> <tex-math notation="LaTeX">$\mathbf {x}$ </tex-math></inline-formula> has probability <inline-formula> <tex-math notation="LaTeX">$h_{u}(\mathbf {x})$ </tex-math></inline-formula> to activate a node <inline-formula> <tex-math notation="LaTeX">$u$ </tex-math></inline-formula> as a seed. LIM is the task of finding a strategy mix under the constraint <inline-formula> <tex-math notation="LaTeX">$\sum _{j} x_{j} \le k$ </tex-math></inline-formula> such that its influence spread is maximized. We adopt the reverse influence sampling (RIS) approach and design scalable algorithms for LIM. We explore two complementary design choices: one algorithm <italic>IMM-PRR</italic> is based on partial coverage on reverse-reachable sets, and the other <italic>IMM-VSN</italic> is based on incorporating virtual strategy nodes. <italic>IMM-PRR</italic> can be applied as a general solution to LIM, and we further improve its efficiency for a large family of models where each strategy independently activates seed nodes. <italic>IMM-VSN</italic> is explicitly designed for the case of independent strategy activation, and it uses virtual nodes to represent strategies to reduce LIM back to the original influence maximization problem. We prove that both <italic>IMM-PRR</italic> and <italic>IMM-VSN</italic> guarantee <inline-formula> <tex-math notation="LaTeX">$1-1/e-\varepsilon $ </tex-math></inline-formula> approximation for small <inline-formula> <tex-math notation="LaTeX">$\varepsilon > 0$ </tex-math></inline-formula>. We further extend LIM to the partitioned budget case where strategies are partitioned into groups, each of which has a separate budget, and show that a minor variation of our algorithms would achieve <inline-formula> <tex-math notation="LaTeX">$1/2 -\varepsilon $ </tex-math></inline-formula> approximation ratio with the same time complexity. Empirically, through extensive tests, we demonstrate that <italic>IMM-VSN</italic> runs faster than <italic>IMM-PRR</italic> and much faster than other baseline algorithms while providing the same level of influence spread. We conclude that <italic>IMM-VSN</italic> is the best one for models with independent strategy activations, while <italic>IMM-PRR</italic> works for general modes without this assumption.

[1]  Cheng Long,et al.  Minimizing Seed Set for Viral Marketing , 2011, 2011 IEEE 11th International Conference on Data Mining.

[2]  Wei Chen,et al.  Efficient influence maximization in social networks , 2009, KDD.

[3]  Laks V. S. Lakshmanan,et al.  Information and Influence Propagation in Social Networks , 2013, Synthesis Lectures on Data Management.

[4]  Yuichi Yoshida,et al.  A Generalization of Submodular Cover via the Diminishing Return Property on the Integer Lattice , 2015, NIPS.

[5]  M. L. Fisher,et al.  An analysis of approximations for maximizing submodular set functions—I , 1978, Math. Program..

[6]  Matthew Richardson,et al.  Mining knowledge-sharing sites for viral marketing , 2002, KDD.

[7]  Paulo Shakarian,et al.  Reducing gang violence through network influence based targeting of social programs , 2014, KDD.

[8]  Xiaokui Xiao,et al.  Influence maximization: near-optimal time complexity meets practical efficiency , 2014, SIGMOD Conference.

[9]  Edith Cohen,et al.  Sketch-based Influence Maximization and Computation: Scaling up with Guarantees , 2014, CIKM.

[10]  Kyomin Jung,et al.  IRIE: Scalable and Robust Influence Maximization in Social Networks , 2011, 2012 IEEE 12th International Conference on Data Mining.

[11]  Amin Karbasi,et al.  Gradient Methods for Submodular Maximization , 2017, NIPS.

[12]  Andreas Krause,et al.  Stochastic Submodular Maximization: The Case of Coverage Functions , 2017, NIPS.

[13]  Joseph Naor,et al.  A Unified Continuous Greedy Algorithm for Submodular Maximization , 2011, 2011 IEEE 52nd Annual Symposium on Foundations of Computer Science.

[14]  Ken-ichi Kawarabayashi,et al.  Optimal Budget Allocation: Theoretical Guarantee and Efficient Algorithm , 2014, ICML.

[15]  Le Song,et al.  Influence Estimation and Maximization in Continuous-Time Diffusion Networks , 2016, ACM Trans. Inf. Syst..

[16]  Andreas Krause,et al.  Guaranteed Non-convex Optimization: Submodular Maximization over Continuous Domains , 2016, AISTATS.

[17]  Christian Borgs,et al.  Maximizing Social Influence in Nearly Optimal Time , 2012, SODA.

[18]  Andreas Krause,et al.  Cost-effective outbreak detection in networks , 2007, KDD '07.

[19]  Michel Minoux,et al.  Accelerated greedy algorithms for maximizing submodular set functions , 1978 .

[20]  Yifei Yuan,et al.  Scalable Influence Maximization in Social Networks under the Linear Threshold Model , 2010, 2010 IEEE International Conference on Data Mining.

[21]  Nicola Barbieri,et al.  Topic-aware social influence propagation models , 2012, Knowledge and Information Systems.

[22]  Wei Chen,et al.  Scalable influence maximization for independent cascade model in large-scale social networks , 2012, Data Mining and Knowledge Discovery.

[23]  Krishna P. Gummadi,et al.  Measurement and analysis of online social networks , 2007, IMC '07.

[24]  Morteza Zadimoghaddam,et al.  How to influence people with partial incentives , 2014, WWW.

[25]  Wei Chen,et al.  An Issue in the Martingale Analysis of the Influence Maximization Algorithm IMM , 2018, CSoNet.

[26]  Jimeng Sun,et al.  Social influence analysis in large-scale networks , 2009, KDD.

[27]  Divyakant Agrawal,et al.  Limiting the spread of misinformation in social networks , 2011, WWW.

[28]  Laks V. S. Lakshmanan,et al.  SIMPATH: An Efficient Algorithm for Influence Maximization under the Linear Threshold Model , 2011, 2011 IEEE 11th International Conference on Data Mining.

[29]  My T. Thai,et al.  Stop-and-Stare: Optimal Sampling Algorithms for Viral Marketing in Billion-scale Networks , 2016, SIGMOD Conference.

[30]  Junsong Yuan,et al.  Online Processing Algorithms for Influence Maximization , 2018, SIGMOD Conference.

[31]  Wei Chen,et al.  Influence Blocking Maximization in Social Networks under the Competitive Linear Threshold Model , 2011, SDM.

[32]  Matthew Richardson,et al.  Mining the network value of customers , 2001, KDD '01.

[33]  Xiaokui Xiao,et al.  Influence Maximization in Near-Linear Time: A Martingale Approach , 2015, SIGMOD Conference.

[34]  Jian Pei,et al.  Continuous Influence Maximization: What Discounts Should We Offer to Social Network Users? , 2016, SIGMOD Conference.

[35]  Laks V. S. Lakshmanan,et al.  From Competition to Complementarity: Comparative Influence Diffusion and Maximization , 2015, Proc. VLDB Endow..

[36]  Éva Tardos,et al.  Maximizing the Spread of Influence through a Social Network , 2015, Theory Comput..

[37]  Weizhong Zhang,et al.  Gradient Method for Continuous Influence Maximization with Budget-Saving Considerations , 2020, AAAI.

[38]  Jinhui Tang,et al.  Online Topic-Aware Influence Maximization , 2015, Proc. VLDB Endow..