Subgraph Counting

This repository contains code for counting subgraph structures for homogeneous graphs (torch_geometric.data.Data). Also contains plotting functionality. The subgraph counting algorithms are used from the python-igraph package. More specifically following algorithm is used:

• VF2 This algorith allows counting subgaphs with vertex and edge constraints. Contrains for vertex and edge matching can be specified by passing vertex/edge color information to the algorithm. If more specific constraints are required, a function can be passed which can evaluate if two vertices or edges are matching.

Setup

Clone repository

git clone https://github.com/stefanjp/SubgraphCounting
cd SubgraphCounting

Add module to PYTHONPATH

export PYTHONPATH=$PYTHONPATH:<PROJECT_ROOT>

where <PROJECT_ROOT> is the project root.

Dependencies

Used following conda channels:

• conda-forge
• pytorch
• pyg

conda install pytorch==1.12.0 torchvision==0.13.0 torchaudio==0.12.0 -c pytorch
conda install pyg -c pyg 
conda install python-igraph matplotlib -c conda-forge
conda install pytorch-lightning -c conda-forge

or create a new environment (might only work on Windows)

conda create --name <env> --file requirements.txt

Code was developed and tested on Windows 11 with following package versions:

• pyg==2.1.0
• python-igraph==0.10.4
• pytorch==1.12.0

Experiments

From the project root directory execute following python file:

python experiments/run_experiments.py

• ZINC-star-3-star-4* experiments
  • loss /test-feature-0 correspons to the MSE of 3-star
  • loss /test-feature-1 correspons to the MSE of 4-star
• ZINC-cycle-3-cycle-6* experiments
  • loss /test-feature-0 correspons to the MSE of 3-cycle
  • loss /test-feature-1 correspons to the MSE of 6-cycle

For the experiments considering node labels, execute following experiment:

python experiments/run_cycle_experiments_node_features.py

The baseline results as well as the model results will be logged in the tb-logs directory that can be visualized with tensorboard.