Metadata-Version: 2.4
Name: fast-graph-gp
Version: 0.1.1
Summary: Graph Random Features for Scalable Gaussian Processes (GRF-GP)
Author-email: Matthew Zhang <matthew.zhang473@gmail.com>
License: MIT
Project-URL: Homepage, https://github.com/MatthewZhang473/GRF-GP
Project-URL: Repository, https://github.com/MatthewZhang473/GRF-GP
Project-URL: Documentation, https://github.com/MatthewZhang473/GRF-GP#readme
Project-URL: Issues, https://github.com/MatthewZhang473/GRF-GP/issues
Keywords: gaussian-processes,graph-kernels,machine-learning,pytorch
Classifier: Development Status :: 4 - Beta
Classifier: Intended Audience :: Science/Research
Classifier: License :: OSI Approved :: MIT License
Classifier: Operating System :: OS Independent
Classifier: Programming Language :: Python :: 3.10
Classifier: Programming Language :: Python :: 3.11
Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
Requires-Python: >=3.8
Description-Content-Type: text/markdown
Requires-Dist: torch>=2.0.0
Requires-Dist: gpytorch>=1.10
Requires-Dist: numpy>=1.21.0
Requires-Dist: scipy>=1.10.0
Requires-Dist: tqdm>=4.65.0

# Graph Random Features for Scalable Gaussian Processes (GRF-GP)



GRF-GP is the package for performing fast Gaussian Process (GP) inference on graphs. Internally, it uses **Graph Random Features** (GRFs) to compute a **unbiased** & **sparse** estimate of a family of well-known graph node kernels. It further uses **path-wise conditioning** to leverage the sparsity of the kernel approximation, enabling you to perform GP model train / inference in $\mathcal{O}(N^{3/2})$ time and $\mathcal{O}(N)$ space complexity.


## Examples

For a detailed example of training and using a Graph GP model, refer to the [example notebook](examples/basic_usage.ipynb).

## Installation

Install GRF-GP via pip:

```bash
pip install grf-gp
```

## Citing Us

If you use GRF-GP, please cite the following papers:

    @article{zhang2025graph,
    title={Graph random features for scalable Gaussian processes},
    author={Zhang, Matthew and Lin, Jihao Andreas and Choromanski, Krzysztof and Weller, Adrian and Turner, Richard E and Reid, Isaac},
    journal={arXiv preprint arXiv:2509.03691},
    year={2025}
    }
