Implementation for paper: Graph Explicit Pooling for Graph-Level Representation Learning

Graph pooling has been increasingly recognized as crucial for Graph Neural Networks (GNNs) to facilitate hierarchical graph representation learning. Existing graph pooling methods commonly consist of two stages: selecting top-ranked nodes and discarding the remaining to construct coarsened graph representations. However, this paper highlights two key issues with these methods:

1) The process of selecting nodes to discard frequently employs additional Graph Convolutional Networks or Multilayer Perceptrons, lacking a thorough evaluation of each node's impact on the final graph representation and subsequent prediction tasks.

2) Current graph pooling methods tend to directly discard the noise segment (dropped) of the graph without accounting for the latent information contained within these elements.

To address the first issue, we introduce a novel Graph explicit Poolng (GrePool) method, which selects nodes by explicitly leveraging the relationships between the nodes and final representation vectors crucial for classification. The second issue is addressed using an extended version of GrePool (i.e., GrePool+), which applies a uniform loss on the discarded nodes. This addition is designed to augment the training process and improve classification accuracy.

conda create -n grepool python=3.10
conda install pytorch==2.1.0 torchvision==0.16.0 torchaudio==2.1.0 pytorch-cuda=11.8 -c pytorch -c nvidia
conda install pyg -c pyg
pip install rdkit-pypi cython
pip install ogb
pip install pyarrow
pip install transformers
pip install configargparse
pip install loguru
pip install wandb
pip install nvidia-ml-py3
pip install tensorboardX
pip install sortedcontainers
pip install pyyaml
pip install pyg_lib torch_scatter torch_sparse torch_cluster torch_spline_conv -f https://data.pyg.org/whl/torch-2.1.0+cu118.html

conda activate grepool
cd gtpooling_basic_final/
# Running grepool topk version using DD dataset
sh run_tu_dd.sh 0
# Running grepool topk version using ogb datasets
sh run_ogb.sh 0
sh run_pcba.sh 0

Supported datasets:

• TuDataset: NCI1, MUTAG, PTC-MR, NCI109, ENZYMES, Mutagenicity, IMDB-M, IMDB-B
• OGB: ogbg-molhiv, ogbg-moltox21, ogbg-moltoxcast, ogbg-molpcba

Datasets mentioned above will be downloaded automatically using PyG's API when running the code.