Influence Maximization with Novelty Decay in Social Networks
Authors: Shanshan Feng, Xuefeng Chen, Gao Cong, Yifeng Zeng, Yeow Meng Chee, Yanping Xiang
AAAI 2014 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We demonstrate the performance of our algorithms on four social networks. Experimental results show that our algorithms can achieve large influence spread efficiently. |
| Researcher Affiliation | Academia | Shanshan Feng,1 Xuefeng Chen,2 Gao Cong,3 1 Interdisciplinary Graduate School, Nanyang Technological University, Singapore, sfeng003@e.ntu.edu.sg 2 University of Electronic Science and Technology of China, China, cxflovechina@gmail.com 3 School of Computer Engineering, Nanyang Technological University, Singapore, gaocong@ntu.edu.sg Yifeng Zeng,4 Yeow Meng Chee,5 Yanping Xiang2 4 School of Computing, Teesside University, UK, Y.Zeng@tees.ac.uk 5 School of Physical and Mathematical Sciences, Nanyang Technological University, Singapore, ymchee@ntu.edu.sg 2 University of Electronic Science and Technology of China, China, xiangyanping@gmail.com |
| Pseudocode | Yes | Algorithm 1: R-Greedy Algorithm with DP. Algorithm 2: Computing σ(S) based on PPND. |
| Open Source Code | No | The paper does not provide any explicit statements about releasing source code or links to a code repository. |
| Open Datasets | Yes | Digg Dataset contains information about stories promoted to the front page of Digg (digg.com) in June 2009 (Lerman and Ghosh 2010). Flickr Dataset contains a friendship graph and a list of favorite marking records from Flickr (www.flickr.com) (Cha, Mislove, and Gummadi 2009). Wiki is a voting network containing all the Wikipedia voting data from the inception of Wikipedia till January 2008. Net PHY is a collection network of papers, extracted from Physics sections in ar Xiv. |
| Dataset Splits | No | The paper uses various datasets (Digg, Flickr, Wiki, Net PHY) but does not provide specific train/validation/test dataset splits, percentages, or sample counts. |
| Hardware Specification | Yes | All methods are implemented in C++ and experiments are conducted on a windows server with 6-core Intel(R) Xeon (R), 2.66 GHz CPU and 24 GB memory. |
| Software Dependencies | No | The paper states, 'All methods are implemented in C++', but does not provide specific version numbers for C++ or any other software dependencies, libraries, or frameworks used. |
| Experiment Setup | Yes | Parameter Setting We set the influence probability Puv of u on v by the weighted cascade policy (Chen, Lu, and Zhang 2012; Liu et al. 2012), i.e., Puv = 1 indegree(v), where indegree(v) is indegree of node v. The expected influencing delay time Tuv of edge uv follows the geometric delay distribution (Chen, Lu, and Zhang 2012). The parameter for geometric distribution is set at 5/(outdegree(v) + 5). For simplicity, the maximum value of Tuv is 15. If the generated delay time Tuv > 15, Tuv is reset as a random integer from 1 to 15. We also try other distributions for Tuv including poisson distribution and uniform distribution, and the experiment results exhibit similar trend for the evaluated techniques. Empirically, threshold is set at = 0.001 and number of paths C is set at C = 5, which achieves a satisfying tradeoff between influence spread and running time in our experiment. |