Graph Ultra-sparsifier

This is the original pytorch implementation of Graph Ultra-sparsifier in the following paper:

[SEMI-SUPERVISED GRAPH ULTRA-SPARSIFIERS USING REWEIGHTED ℓ1 OPTIMIZATION, ICASSP 2023]

Datasets

Here using Cora as an example.

Pretrained step

python pretrain.py

Note that different datasets should correspond to different parameter settings in the pretrained step.

Sparse training step

python sparse_train.py

Note that you may need to change some parameter settings during the sparse training.

Appendix

Some supplementary materials are included in the Appendix.pdf.