SLiCE
Self-Supervised Learning of Contextual Embeddings for Link Prediction in Heterogeneous Networks
Dataset details:
- We use four public benchmark datasets covering multiple applications: e-commerce (Amazon), academic graph (DBLP), knowledge graphs (Freebase) and social networks (Twitter). Amazon and Twitter data came from https://github.com/THUDM/GATNE. Freebase data came from https://github.com/malllabiisc/CompGCN. DBLP data came from https://github.com/Jhy1993/HAN.
- We introduce a new knowledge graph from the publicly available real-world Medical Information Mart for Intensive Care III (MIMIC III) dataset in healthcare domain. https://mimic.physionet.org/
- We also introduce a new knowledge graph from the publicly available Intrusion detection evalution dataset (ISCXIDS2012) https://www.unb.ca/cic/datasets/ids.html
- Note: Relationship IDs were converted to 1-based indexing if they were previously 0-based
Install instructions:
- Dependencies: Python 3.6, PyTorch 1.4.0 w/ CUDA 9.2, Pytorch Geometric
- The specific Pytorch Geometric wheels we use are included in the repo for convenience in the 'wheels' directory
conda create -n slice python=3.6
conda activate slice
pip install -r requirements.txt
Training:
python main.py \
--data_name 'amazon_s' \
--data_path 'data' \
--outdir 'output/amazon_s' \
--pretrained_embeddings 'data/amazon_s/amazon_s.emd' \
--n_epochs 10 \
--n_layers 4 \
--n_heads 4 \
--gcn_option 'no_gcn' \
--node_edge_composition_func 'mult' \
--ft_input_option 'last4_cat' \
--path_option 'shortest' \
--ft_n_epochs 10 \
--num_walks_per_node 1 \
--max_length 6 \
--walk_type 'dfs' \
--is_pre_trained
Citation:
Please cite the following paper if you use this code in your work.
@inproceedings{wang2020self,
title={Self-Supervised Learning of Contextual Embeddings for Link Prediction in Heterogeneous Networks},
author={Wang, Ping and Agarwal, Khushbu and Ham, Colby and Choudhury, Sutanay and Reddy, Chandan K},
booktitle={Proceedings of The Web Conference 2021},
year={2021}
}
Notice
This material was prepared as an account of work sponsored by an agency of the United States Government. Neither the United States Government nor the United States Department of Energy, nor Battelle, nor any of their employees, nor any jurisdiction or organization that has cooperated in the development of these materials, makes any warranty, express or implied, or assumes any legal liability or responsibility for the accuracy, completeness, or usefulness or any information, apparatus, product, software, or process disclosed, or represents that its use would not infringe privately owned rights. Reference herein to any specific commercial product, process, or service by trade name, trademark, manufacturer, or otherwise does not necessarily constitute or imply its endorsement, recommendation, or favoring by the United States Government or any agency thereof, or Battelle Memorial Institute. The views and opinions of authors expressed herein do not necessarily state or reflect those of the United States Government or any agency thereof.
PACIFIC NORTHWEST NATIONAL LABORATORY
operated by
BATTELLE
for the
UNITED STATES DEPARTMENT OF ENERGY
under Contract DE-AC05-76RL01830
License
Released under the 3-Clause BSD license (see License.md)
51 Dec 3, 2022
27 Dec 2, 2022
88 Dec 25, 2022
19 Sep 29, 2022
306 Dec 2, 2022
11 Oct 20, 2022
106 Jan 6, 2023
4 Jun 9, 2021
37 Dec 6, 2022
Twitter:
nohup python3 main.py --base_embedding_dim=128 --batch_size=128 --beam_width=4 --checkpoint=15 --d_ff=512 --d_k=64 --d_model=128 --d_v=64 --data_name='twitter' --emb_dir=None --false_edge_gen='double' --ft_batch_size=100 --ft_checkpoint=500 --ft_d_ff=512 --ft_drop_rate=0.1 --ft_input_option='last4_cat' --ft_layer='ffn' --ft_lr=5e-05 --ft_n_epochs=10 --gcn_option='no_gcn' --get_bert_encoder_embeddings=False --lr=0.0001 --max_length=12 --n_epochs=10 --n_heads=4 --n_layers=4 --node_edge_composition_func='mult' --num_gcn_layers=2 --num_walks_per_node=10 --path_option='shortest' --pretrained_embeddings='./embed/twitter/twitter.emd' --outdir='./output/twitter/' --pretrained_method='node2vec' --walk_type='dfs' 2>&1 &
58 Dec 31, 2022
50 Dec 3, 2022
105 Dec 23, 2022
51 Dec 22, 2022
48 Jan 2, 2023
71 Oct 25, 2022
8 Jun 23, 2021
55 Nov 9, 2022
23 Dec 26, 2022
19 Oct 21, 2022