RecSys'16 - RBM and DBM model in recommender systems
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
EVAL
data
model
pretrain
.DS_Store
README.txt
dwr_sampling.m
eval_recall.m
infer_dbm.m
infer_dwr.m
infer_rbm.m
inference.m
init_delicious.m
init_lastfm.m
main.m
makebatches.m
meval_noise.m
mf_dbm.m
pp.m
pp_2D.m
pre_rbm_l1.m
pre_rbm_l2.m
rankFriends.m
rbm.m
rbm_sampling.m
sampling.m
sigmoid.m
sodbm.m
sodwr.m

README.txt

SoialRBM: Social Restricted Boltzmann Machine
------------------------------------------

INTRODUCTION: 

- This is an implementation in Matlab for the paper "Representation Learning for Homophilic Preferences" (Nguyen & Lauw, RecSys 2016)


The folder includes:
- Data: Delicious and LastFM, with 10-samples included for training and testing.
- Eval: store results from the inference stage.
- Model: store corresponding model for each specific algorithm.
- Pretrain: pretraining RBM for each layer in deep Boltzmann machine (DBM) algorithm.


*** Running from Matlab with "main" function:
main(x1,x2,x3,x4,x5)

with arguments in order below:
- x1: algorithm selection as below:
	+ 1: RBM (for item only)
	+ 2: Social Wing Model (user network + item adoptions) ("SocialRBM-Wing" as Section 4)
	+ 3: Random Social Wing Model (with RANDOM user-network)
	+ 4: Social Deep Model ("SocialRBM-Deep" as Section 5)
	+ 5: Random Social Deep Model (with RANDOM user-network)
- x2: data selection (1-Lastfm; 2-Delicious)
- x3: data-sample selection (from 1 to 10)
- x4(optional): the number of hidden units (default: 100)
- x5(optional): using GPU or not (default: 0)


Examples:
- Running "Social-Wing" algorithm on "Delicious" dataset with "100" hidden units on Sample "3" (with GPU).
main(2,2,3,100,1)


HOW TO CITE:

If you use SoialRBM in your research, please cite the paper with the bibtex format below:

@inproceedings{nguyen2016representation,
  title={Representation learning for homophilic preferences},
  author={Nguyen, Trong T and Lauw, Hady W},
  booktitle={Proceedings of the 10th ACM Conference on Recommender Systems},
  pages={317--324},
  year={2016},
  organization={ACM}
}