The codes are for our paper "Unsupervised Learning of Graph Hierarchical Abstractions with Differentiable Coarsening and Optimal Transport" in AAAI 2021.
The codes are built on Pytorch Geometric library. Please configure your environment as PyG requires (e.g. Pytorch version>=1.2.0).
The main file is "main_opt2.py". The following is an example script for running:
python -u ./src/main_opt2.py --eps=1.0 --lr=0.001 --opt_iters=10 --no_extra_mlp --train --epochs=100