Graph Gaussian Process (GGP)

The code and data in this repository accompany the paper `Bayesian Semi-supervised Learning with Graph Gaussian Processes'

@inproceedings{ng2018gaussian,
  title={Bayesian semi-supervised learning with graph Gaussian processes},
  author={Ng, Yin Cheng and Colombo, Nicolo and Silva, Ricardo},
  booktitle={Advances in Neural Information Processing Systems},
  year={2018}
}

The code depends on a branch of GPflow located here.

To run the graph-based semi-supervised learning experiment, execute the following command:

python ssl_exp.py [name of the data set] [random seed]
valid options for the name of the data set are: cora, citeseer or pubmed
valid options for the random seed: any integer

To run the active learning experiment, execute the following command:

python al_exp.py [name of the data set] [random seed]
valid options for the name of the data set are: cora or citeseer
valid options for the random seed: any integer