Hyperbolic SVM Relaxation

SDP and Moment Relaxation of Hyperbolic SVM.

For details, see our preprint: https://arxiv.org/abs/2405.17198.

Environment

Install the following packages

# create a virtual env (recommend python >= 3.8)
python -m venv hsvm_relax
source ./hsvm_relax/bin/activate

# install packages
pip install matplotlib scikit-learn toml mosek cvxpy SumOfSquares

mosek is a commercial solver. One can get an academic license for free from https://www.mosek.com/products/academic-licenses/.

Running Instructions

to reproduce results, we first specify the relative paths with respect to the root of the project in config.toml:

• data_dir: path to read data;
• model_dir: path to store model parameters;
• result_dir: path to dump Kfold train test results.

along with the associated tags. For experiment tag, see the following snippet also in config.toml

["exp"]
data_dir='data/'
model_dir='model/'
result_dir='result/'

To run experiments, simply follow the bash scripts in commands/. One may directly run the bash scripts (if with access to a Slurm system) or adapt accordingly.

One example is as follows: use moment relaxation with $C = 10.0$ on krumsiek dataset with a 5 fold train test split (default) can be called by

# run the following at root
python src/train.py --data krumsiek --model moment --C 10. --verbose --dump

where --verbose indicates printing the interior point progress summary and --dump indicates saving the trained model parameters. See src/train.py for a full lists of parameters.

Acknowledgement

If you find this code useful, please consider citing our preprint:

@article{yang2024convex,
  title={Convex Relaxation for Solving Large-Margin Classifiers in Hyperbolic Space},
  author={Yang, Sheng and Liu, Peihan and Pehlevan, Cengiz},
  journal={arXiv preprint arXiv:2405.17198},
  year={2024}
}