Fake News Mitigation via Point Process Based Intervention

Authors: Mehrdad Farajtabar, Jiachen Yang, Xiaojing Ye, Huan Xu, Rakshit Trivedi, Elias Khalil, Shuang Li, Le Song, Hongyuan Zha

ICML 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Our method shows promising performance in real-time intervention experiments on a Twitter network to mitigate a surrogate fake news campaign, and outperforms alternatives on synthetic datasets.
Researcher Affiliation Academia 1School of Computational Science and Engineering, Georgia Tech. 2Department of Mathematics and Statistics, Georgia State University. 3School of Industrial and Systems Engineering, Georgia Tech..
Pseudocode Yes Algorithm 1 LSTD policy iteration in point processes; Algorithm 2 Real-time fake news mitigation
Open Source Code No The paper does not provide concrete access to source code for the methodology described.
Open Datasets No The paper describes using synthetically generated networks and accumulating a network from Twitter data ('accumulated a network of 1894 real users with 23407 directed edges in total'), but does not provide concrete access information (link, DOI, or formal citation for a publicly available dataset) for these datasets.
Dataset Splits No The paper mentions using '1000 randomly sampled states' for the LSTD algorithm and simulating '50 times' for policy evaluation, and using 'historical data to learn the network parameters' for real experiments, but it does not provide specific train/validation/test dataset split information.
Hardware Specification No The paper does not provide specific hardware details such as GPU/CPU models or other computer specifications used for running experiments.
Software Dependencies No The paper does not provide specific ancillary software details with version numbers (e.g., library or solver names and versions) needed to replicate the experiment.
Experiment Setup Yes Endogenous intensity coefficients were set as aij U[0, 0.5]. To mimic real world networks, sparsity was set to 0.02, i.e., each edge was kept with probability 0.02. The influence matrix was scaled appropriately such that the spectral radius is a random number smaller than one to ensure the stability of the process. The Hawkes kernel parameter was set to ! = 1, which means loosing roughly 63 % of influence after 1 unit of time (minutes, hours, etc). Both fake news and mitigation processes obey these network settings. Each stage has length of T = 1. The discount factor was set to = 0.7. For determining features, we set L = 2 and we choose f = T for simplicity. The upper bound for the intervention intensity was chosen by i U[0, 0.5]. The price of each person was ck i = 1, and the total budget at stage k was randomly generated as Ck (n U[0, 0.5]).