This is the source code for paper Influence without Authority: Maximizing Information Coverage in Hypergraphs [SDM'23].
python>=3.3.7
multiprocessing
joblib
tqdm
This code was tested on Windows and Linux.
python InfDis.py --data=email
--data
: contact_primary, contact, email, w3cemail, geology, history, flickr, dblp, stackoverflow
--probability
: independent propagation probability, default=0.01
the seeds of all algorithms will be saved to the data path, e.g., "./data/email/".
python Evaluation.py --data=email --probability=0.01 --num_mcmc=100 --method=InfDis
--data
: contact_primary, contact, email, w3cemail, geology, history, flickr, dblp, stackoverflow
--probability
: independent propagation probability, default=0.01
--earlystopping
: the number of steps of independent cascades
--num_mcmc
: the number of mcmc simulations, default=100
--method
: InfDis, Degree, Between, HyperRank
the results will be saved to a fixed path, e.g., "./result/".
Multiprocessing is enabled by default, and num_mcmc should be larger than the number of CPU cores.
please unzip "data.zip", and run:
./run_demo.bat
./run_demo.sh
@inproceedings{li2023influence,
title={Influence without Authority: Maximizing Information Coverage in Hypergraphs},
author={Li, Peiyan and Wang, Honglian and Li, Kai and Bohm, Christian},
booktitle={Proceedings of the 2023 SIAM International Conference on Data Mining (SDM)},
pages={10--18},
year={2023},
organization={SIAM}
}