Learning-Hypergraphs-From-Signals-With-Dual-Smoothness-Prior

This is the repo for our ICASSP 2023 paper: Learning Hypergraphs From Signals With Dual Smoothness Prior.

Recommend Environment:

python 3.7.10

pytorch 1.5.1

Running Experiments:

python main.py

There are four hyperparameters that may need to be fine-tuned for different datasets: alpha, beta, step_size, and threshold.

In our experiments, we conducted a grid search for alpha and beta, ranging from 1e-3 to 1e+3, for step_size, ranging from 1e-4 to 1, and for threshold, ranging from 1e-4 to 5e-1.

Citation

If you use this code, please cite our paper:

@inproceedings{tang2023learning,
  title={Learning Hypergraphs From Signals With Dual Smoothness Prior},
  author={Tang, Bohan and Chen, Siheng and Dong, Xiaowen},
  booktitle={ICASSP 2023-2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)},
  pages={1--5},
  year={2023},
  organization={IEEE}
}