Code for AAAI2018 paper "Collaborative Filtering with Social Exposure: A Modular Approach to Social Recommendation"
Contact: Menghan Wang (wangmengh@zju.edu.cn)
-
gcc 5.4.0
-
gsl 2.2
-
openMP (If you don't want to use it, you can delete all the lines look like "#pragma omp parallel for..." in the "socialexpo.cpp")
-
(optional) openBLAS
A data sample is provided. (the line numbers of files below are used as the index to users or items, starting from 0.)
-
user file: total_rated_items [space] Item1:count [space] Item2:count ...
-
item file: total_rated_users [space] User1:count [space] User2:count ...
-
social file: friend_sum [space] Friend1:1 [space] Friend2:1 ...
Note that the SERec is new name of our model, in the code it is named "s_expo".
-
Revise the lib path in Makefile to meet your settings.
-
Tune the parameters in run.sh. As there are too many paramters, you may need to tune some parameters in the main.cpp.
-
compile the code and run with "run.sh".
-
The meaning of the parameter "version":
Verison 1 means exposures are computed purely based on popularity, which is equivalent to "ExpoMF" (Our c++ version is much faster).
Verison 2 means "social boosting"
Verison 3 means "social regularization"
Please cite our paper if it is helpful to your research:
@inproceedings{DBLP:conf/aaai/WangZYZ18,
author = {Menghan Wang and
Xiaolin Zheng and
Yang Yang and
Kun Zhang},
title = {Collaborative Filtering With Social Exposure: {A} Modular Approach
to Social Recommendation},
booktitle = {Proceedings of the Thirty-Second {AAAI} Conference on Artificial Intelligence,
New Orleans, Louisiana, USA, February 2-7, 2018},
year = {2018},
crossref = {DBLP:conf/aaai/2018},
url = {https://www.aaai.org/ocs/index.php/AAAI/AAAI18/paper/view/16058},
timestamp = {Thu, 03 May 2018 17:03:19 +0200},
biburl = {https://dblp.org/rec/bib/conf/aaai/WangZYZ18},
bibsource = {dblp computer science bibliography, https://dblp.org}
}