PSI

Official Implementation of 'Partial Subgraph InfoMax (PSI)' from 'Models and Benchmarks for Representation Learning of Partially Observed Subgraphs', 31st ACM International Conference on Information and Knowledge Management (CIKM 2022, Short Papers Track).

BibTeX

TBA

Installation

bash PSI/install.sh ${CUDA, optional, default is cu102.}

If you have any trouble installing PyTorch Geometric, please install PyG's dependencies manually.
Codes are tested with python 3.7.9 and nvidia/cuda:10.2-cudnn8-devel-ubuntu18.04 image.
PYG's FAQ might be helpful.

Basics

The main train/test code is in PSI/main.py.
If you want to see hyperparameter settings, refer to PSI/args.yaml and PSI/arguments.py.

Run

python -u PSI/main.py \
    --dataset-name FNTN \
    --custom-key BIE2D2F64-ISI-X-GB-PGA \
    --gpu-ids 0 \
    --dataset-path /mnt/nas2/GNN-DATA/

GPU Setting

There are three arguments for GPU settings (--num-gpus-total, --num-gpus-to-use, --gpu-ids). Default values are from the author's machine, so we recommend you modify these values from PSI/args.yaml or by the command line.

--num-gpus-total (default 4): The total number of GPUs in your machine.
--num-gpus-to-use (default 1): The number of GPUs you want to use.
--gpu-ids (default: [0]): The ids of GPUs you want to use.

Datasets

Names (`--dataset-name`)

Dataset	`--dataset-name`
FNTN	FNTN
EM-User	EMUser
HPO-Metab	HPOMetab

Path (`--dataset-path`)

Download datasets and put them into the specific path (--dataset-path).

root@5b592ce:~$ ls /mnt/nas2/GNN-DATA/
EMUSER  FNTN  HPOMETAB

Models (`--custom-key`)

Type	FNTN	EMUser & HPOMetab
PS-DGI	BISAGE-SHORT-DGI-X-GB-PGA	SAGE-SHORT-DGI-X-GB
PS-InfoGraph	BISAGE-SHORT-ISI-X-GB-PGA	SAGE-SHORT-ISI-X-GB
PS-MVGRL	BISAGE-SHORT-MVGRL-X-GB-PGA	SAGE-SHORT-MVGRL-X-GB
PS-GraphCL	BISAGE-SHORT-GRAPHCL3-X-GB-PGA	SAGE-SHORT-GRAPHCL3FB-X-GB (only for HPOMetab)
k-hop PSI	BIE2D2F64-X-PGA	E2D2F64-X
k-hop PSI + PS-DGI	BIE2D2F64-DGI-X-GB-PGA	E2D2F64-DGI-X-GB
k-hop PSI + PS-InfoGraph	BIE2D2F64-ISI-X-GB-PGA	E2D2F64-ISI-X-GB

Other Hyperparameters

See PSI/args.yaml or run $ python PSI/main.py --help.

About

[CIKM 2022 Short] Models and Benchmarks for Representation Learning of Partially Observed Subgraphs

Languages

Language:Python 99.7%Language:Shell 0.3%

PSI