Running the experiments

Dataset used in the paper

Data Format:

• temporal graph: Dynamic.txt: u, i, ts , edge feature
• node feature: Node_Features.txt: u, node features

Download

• Social Data:
  • Download Link
  • Download to the folder 'social_data/'
• DPPIN：
  • Download Link
  • Download to the folder 'dppin_data/'

Requirements

• Python >= 3.7
• Package requirements

pandas
torch == 1.8.1
tqdm == 4.59.0
numpy == 1.20.1
scikit_learn == 1.0.2
psutil == 5.8.0
jsonlines == 2.0.0
pytorch_geometric == 1.7.0

How to run code

cd Temp-GFSM
python3 main.py --dir 'social_data/' --batch_size 10 --k_shot 3 --k_query 3 --n_way 3 --num_task 20 --update_step 5 --nhid 32 --update_lr 0.001 --device 1 --labels 

Arguments in `main.py` need to be specified for different datasets:
* Social Data:
   * --labels 'dblp_ct1_1,dblp_ct1_0,facebook_ct1_1,facebook_ct1_0,tumblr_ct1_1,tumblr_ct1_0,highschool_ct1_1,highschool_ct1_0,infectious_ct1_1,infectious_ct1_0,mit_ct1_1,mit_ct1_0'
   * --logdir 'logs-sd-b1-'
   * --output_file''output-sd-'
   * --total_sample_g {'dblp_ct1_1':755,'dblp_ct1_0':755,'facebook_ct1_1':995,'facebook_ct1_0':995,'highschool_ct1_1':179,'highschool_ct1_0':179, 'infectious_ct1_1':199, 'infectious_ct1_0':199,'mit_ct1_1':79,'mit_ct1_0':79,'tumblr_ct1_1':373, 'tumblr_ct1_0':373}
* DPPIN Data:
   * --labels 'Uetz,Yu,Babu,Breitkreutz,Gavin,Hazbun,Ho,Ito,Krogan_LCMS,Krogan_MALDI,Lambert,Tarassov'
   * --logdir 'logs-ppin-b1-'
   * --output_file''output-ppin-'
   * --total_sample_g 11

This repository is for the KDD' 2022 paper "Meta-Learned Metrics over Multi-Evolution Temporal Graphs" (Link) .

Functionality

Temp-GFSM first models temporal graphs for multiple dynamic evolution pattern, then it learns the accurate and adaptive metrics over them via the representation learning techniques.

Reference

If you use the materials from this repositiory, please refer to our paper.

@inproceedings{DBLP:conf/kdd/FuFMTH22,
  author    = {Dongqi Fu and
               Liri Fang and
               Ross Maciejewski and
               Vetle I. Torvik and
               Jingrui He},
  title     = {Meta-Learned Metrics over Multi-Evolution Temporal Graphs},
  booktitle = {{KDD} '22: The 28th {ACM} {SIGKDD} Conference on Knowledge Discovery
               and Data Mining, Washington, DC, USA, August 14 - 18, 2022},
  pages     = {367--377},
  publisher = {{ACM}},
  year      = {2022}
}