Skip to content
/ Zebra Public

Code repository for the VLDB2023 paper "Zebra: When Temporal Graph Neural Networks Meet Temporal Personalized PageRank".

Notifications You must be signed in to change notification settings

LuckyLYM/Zebra

Folders and files

NameName
Last commit message
Last commit date

Latest commit

8e72d65 · Apr 14, 2023

History

22 Commits
Feb 12, 2023
Apr 14, 2023
Feb 12, 2023
Apr 14, 2023
Feb 24, 2023
Feb 22, 2023
Apr 14, 2023

Repository files navigation

Zebra: When Temporal Graph Neural Networks Meet Temporal Personalized PageRank

Dataset

6 datasets were used in this paper:

Preprocessing

If edge features or nodes features are absent, they will be replaced by a vector of zeros. Example usage:

python utils/preprocess_data.py --data wikipedia --bipartite
python utils/preprocess_custom_data.py --data superuser

Usage

Optional arguments:
    --data                  Dataset name
    --bs                    Batch size
    --n_head                Number of attention heads used in neighborhood aggregation
    --n_epoch               Number of training epochs
    --n_layer               Number of network layers
    --lr                    Learning rate
    --gpu                   GPU id
    --patience              Patience for early stopping
    --enable_random         Use random seeds
    --topk                  Top-k threshold
    --tppr_strategy         Strategy used for answering top-k T-PPR query [streaming|pruning]
    --alpha_list            Alpha values used in T-PPR metrics
    --beta_list             Beta values used in T-PPR metrics
    
Example usage:
    python train.py --n_epoch 50 --bs 200 --data wikipedia --enable_random  --tppr_strategy streaming  --topk 20 --alpha_list 0.1 0.1 --beta_list 0.5 0.95 --gpu 0

About

Code repository for the VLDB2023 paper "Zebra: When Temporal Graph Neural Networks Meet Temporal Personalized PageRank".

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages