#### DMCA

## Sketch-based influence maximization and computation: Scaling up with guarantees (2014)

Venue: | In International Conference on Information and Knowledge Management (ICIKM |

Citations: | 4 - 0 self |

### Citations

989 | Maximizing the spread of influence through a social network
- Kempe, Kleinberg, et al.
- 2003
(Show Context)
Citation Context ...pped for each directed edge (u, v). Accordingly, the edge can be either live, meaning that once u is infected, v is also infected, or null. This model was formalized in a seminal work by Kempe et al. =-=[19]-=- and is based on earlier studies by Goldenberg et al. [14]. Note that each direction of an undirected edge {u, v} may have its own independent random variable, since influence is not necessarily symme... |

750 |
An analysis of approximations for maximizing submodular set functions - ii
- Fisher, Nemhauser, et al.
- 1978
(Show Context)
Citation Context ...set and iteratively adds to S the node with maximum marginal gain in influence (relative to current seed set). Since our objective is monotone and submodular, a classical result from Nemhauser et al. =-=[21]-=- implies that the influence of the greedy solution with s seeds is at least 1 − (1 − 1/s)s ≥ 63% of the best possible for any seed set of the same size. From Feige’s inapproximability result, this is ... |

368 | Domingos,Mining knowledge-sharing sites for viral marketing
- Richardson, P
- 2002
(Show Context)
Citation Context ...s can have one or multiple sources (or seeds) and spread from infected nodes to neighbors through the link structure. A motivating application for the study of influence is viral marketing strategies =-=[14, 23]-=-, in which the influence of a set S of people in a social network is the number of adoptions triggered if we give S free copies of a product. The problem also has important applications beyond social ... |

322 | Cost-effective outbreak detection in networks
- LESKOVEC, KRAUSE, et al.
- 2007
(Show Context)
Citation Context ...bution of each node requires a directed reachability computation in each instance (of which there can be hundreds). Several performance improvements to Greedy have thus been proposed. Leskovec et al. =-=[20]-=- proposed CELF, which are “lazy” evaluations of the marginal contribution, performed only when a node is a candidate for the highest marginal contribution. Chen et al. [6] took a different approach, u... |

281 |
Talk of the Network: A Complex Systems Look at the Underlying Process of Word-of-Mouth. Marketing Letters
- Goldenberg, Libai, et al.
- 2001
(Show Context)
Citation Context ...s can have one or multiple sources (or seeds) and spread from infected nodes to neighbors through the link structure. A motivating application for the study of influence is viral marketing strategies =-=[14, 23]-=-, in which the influence of a set S of people in a social network is the number of adoptions triggered if we give S free copies of a product. The problem also has important applications beyond social ... |

268 | The WebGraph framework I: Compression techniques
- Boldi, Vigna
- 2004
(Show Context)
Citation Context ...res (2.90GHz, 8× 64 kiB L1, 8 × 256 kiB, and 20MiB L3 cache), but all runs are sequential for consistency. We ran our experiments on benchmark networks available as part of the SNAP [24] and WebGraph =-=[2]-=- projects. More specifically, we test social (Epinions, Slashdot, Gowalla, TwitterFollowers, LiveJournal, Orkut, Friendster, Twitter), collaboration (AstroPh), and web (Slovakia, Slovakia>) networks. ... |

219 |
A threshold of lnn for approximating set cover
- Feige
- 1998
(Show Context)
Citation Context ...e function is deterministic (but the number s of seeds is a parameter), the problem encodes the classic Max Cover problem and therefore is NP-hard [19]. Moreover, an inapproximability result of Feige =-=[13]-=- 1 ar X iv :1 40 8. 62 82 v1s[ cs .D S]s2 6 A ugs20 14 implies that any algorithm that can guarantee a solution that is at least (1− 1/e+ ) times the optimum is likely to scale poorly with the number... |

197 | Efficient influence maximization in social networks
- CHEN, WANG, et al.
- 2009
(Show Context)
Citation Context ...n proposed. Leskovec et al. [20] proposed CELF, which are “lazy” evaluations of the marginal contribution, performed only when a node is a candidate for the highest marginal contribution. Chen et al. =-=[6]-=- took a different approach, using the reachability sketches of Cohen [7] to speed up the reevaluation of the marginal contribution of all nodes. While effective, even with these and other acceleration... |

183 | Scalable influence maximization for prevalent viral marketing in large-scale social networks
- CHEN, WANG, et al.
- 2010
(Show Context)
Citation Context ...82 v1s[ cs .D S]s2 6 A ugs20 14 implies that any algorithm that can guarantee a solution that is at least (1− 1/e+ ) times the optimum is likely to scale poorly with the number of seeds. Chen et al. =-=[5]-=- showed that computing the exact influence of a single seed in the binary IC model, even when edge probabilities are p = 0.5, is #P hard [5]. Using simulations, the objective studied by Kempe et al. [... |

158 | Size-estimation framework with applications to transitive closure and reachability
- Cohen
- 1997
(Show Context)
Citation Context ...ions of the marginal contribution, performed only when a node is a candidate for the highest marginal contribution. Chen et al. [6] took a different approach, using the reachability sketches of Cohen =-=[7]-=- to speed up the reevaluation of the marginal contribution of all nodes. While effective, even with these and other accelerations [17, 22], the best current implementations of Greedy do not scale to n... |

116 | Inferring networks of diffusion and influence.
- Gomez-Rodriguez, Leskovec, et al.
- 2010
(Show Context)
Citation Context ...l. The ability to compute influence with respect to an arbitrary set of propagation instances has significant advantages, as it is useful for instances generated from traces or by more complex models =-=[16, 1]-=-, which exhibit correlations between edges that cannot be captured by the simplified IC model [15]. Moreover, the average behavior of a probabilistic model on a small set of instances captures its “ty... |

56 | Uncovering the Temporal Dynamics of Diffusion Networks.
- Gomez-Rodriguez, Balduzzi, et al.
- 2011
(Show Context)
Citation Context ...nificant advantages, as it is useful for instances generated from traces or by more complex models [16, 1], which exhibit correlations between edges that cannot be captured by the simplified IC model =-=[15]-=-. Moreover, the average behavior of a probabilistic model on a small set of instances captures its “typical” behavior, which is often more relevant than the expected value when the variance is very hi... |

35 | CELF++: optimizing the greedy algorithm for influence maximization in social networks.
- Goyal, Lu, et al.
- 2011
(Show Context)
Citation Context ...ook a different approach, using the reachability sketches of Cohen [7] to speed up the reevaluation of the marginal contribution of all nodes. While effective, even with these and other accelerations =-=[17, 22]-=-, the best current implementations of Greedy do not scale to networks beyond 106 edges [5], which are quite small by modern standards. To support massive graphs, several studies proposed algorithms sp... |

30 | Summarizing data using bottom-k sketches.
- Cohen, Kaplan
- 2007
(Show Context)
Citation Context ...ts its influence “coverage” across ` instances; we call this its combined reachability set. The combined reachability sketch of a node, precisely defined in Section 3, is the bottom-k min-hash sketch =-=[10, 8]-=- of the combined reachability set of the node. This generalizes the reachability sketches of Cohen [7], which are defined for a single instance. The parameter k is a small constant that determines the... |

23 | Scalable influence estimation in continuous-time diffusion networks. - Du, Song, et al. - 2013 |

22 |
Network Analysis Project. http://snap.stanford.edu/index.html
- Stanford
(Show Context)
Citation Context ... Each CPU has 8 cores (2.90GHz, 8× 64 kiB L1, 8 × 256 kiB, and 20MiB L3 cache), but all runs are sequential for consistency. We ran our experiments on benchmark networks available as part of the SNAP =-=[24]-=- and WebGraph [2] projects. More specifically, we test social (Epinions, Slashdot, Gowalla, TwitterFollowers, LiveJournal, Orkut, Friendster, Twitter), collaboration (AstroPh), and web (Slovakia, Slov... |

18 | Maximizing social influence in nearly optimal time.
- Borgs, Brautbar, et al.
- 2014
(Show Context)
Citation Context ...osed algorithms specific to the IC model, which work directly with the edge probabilities instead of with simulations and thus can not be reliably applied to a set of arbitrary instances. Borg et al. =-=[3]-=- recently proposed an algorithm based on reverse reachability searches from sampled nodes, similar in spirit to the approach used for reachability sketching [7]. Their algorithm provides theoretical g... |

16 | Time-critical influence maximization in social networks with time-delayed diffusion - Chen, Lu, et al. |

13 | Irie: Scalable and robust influence maximization in social networks
- Jung, Heo, et al.
- 2012
(Show Context)
Citation Context ...adding the next highest degree node. MIA [5] converts the binary IC sampling probabilities pe to deterministic edge weights and works essentially with one deterministic instance. IRIE, by Jung et al. =-=[18]-=-, is a heuristic approximation of greedy addition of seed nodes, and has the best performance we are aware of for an algorithm that produces a sequence of seed nodes. In each step, the probability of ... |

12 | Influence maximization: near-optimal time complexity meets practical efficiency.
- Tang, Xiao, et al.
- 2014
(Show Context)
Citation Context ...ed for reachability sketching [7]. Their algorithm provides theoretical guarantees on the approximation quality and has good asymptotic performance, but large “constants.” Very recently, Tang et. al. =-=[25]-=- developed TIM, which engineers the (mostly theoretical) algorithm of Borgs et al. [3] to obtain a scalable implementation with guarantees. A significant drawback of this approach is that it only work... |

9 | Leveraging discarded samples for tighter estimation of multiple-set aggregates
- Cohen, Kaplan
- 2009
(Show Context)
Citation Context ...re built in O(k ∑ i m(i)) total time. The influence of a set S ⊆ V can then be approximated from the sketches of the nodes in S. The oracle applies the union cardinality estimator of Cohen and Kaplan =-=[11]-=- to estimate the union of the influence sets of the seed nodes. The query runs in time O(|S|k log |S|) and unbiasedly with a well-concentrated relative error of = 1/ √ k. While preprocessing depends... |

7 | All-distances sketches, revisited: HIP estimators for massive graphs analysis
- Cohen
- 2014
(Show Context)
Citation Context ... randomly selected from instances j for which the pair (v, j) does not have a permutation rank of in or less (independently for each node). One can show that this can only improve estimation accuracy =-=[8]-=-. Only the first min{k, `}n positions can be included in combined reachability sketches of nodes. When estimating influence, we can convert permutation ranks to random ranks using the exponential dist... |

6 | Scalable similarity estimation in social networks: Closeness, node labels, and random edge length - Cohen, Delling, et al. - 2013 |

5 | Trace complexity of network inference
- Abrahao, Chierichetti, et al.
- 2013
(Show Context)
Citation Context ...l. The ability to compute influence with respect to an arbitrary set of propagation instances has significant advantages, as it is useful for instances generated from traces or by more complex models =-=[16, 1]-=-, which exhibit correlations between edges that cannot be captured by the simplified IC model [15]. Moreover, the average behavior of a probabilistic model on a small set of instances captures its “ty... |

2 | Fast and accurate influence maximization on large networks with pruned monte-carlo simulations
- Ohsaka, Akiba, et al.
- 2014
(Show Context)
Citation Context ...ook a different approach, using the reachability sketches of Cohen [7] to speed up the reevaluation of the marginal contribution of all nodes. While effective, even with these and other accelerations =-=[17, 22]-=-, the best current implementations of Greedy do not scale to networks beyond 106 edges [5], which are quite small by modern standards. To support massive graphs, several studies proposed algorithms sp... |