qpytorch 0.1

An implementation of Q-Exponential Processes in Pytorch based on GPyTorch

Q^ⓔPyTorch

---

Q^ⓔPyTorch is a Q-exponential process library implemented using PyTorch built on GPyTorch. Q^ⓔPyTorch is designed for creating scalable, flexible, and modular Q-exponential process models with ease.

Internally, Q^ⓔPyTorch differs from many existing approaches to QEP inference by performing most inference operations using numerical linear algebra techniques like preconditioned conjugate gradients. Implementing a scalable QEP method is as simple as providing a matrix multiplication routine with the kernel matrix and its derivative via our LinearOperator interface, or by composing many of our already existing LinearOperators. This allows not only for easy implementation of popular scalable QEP techniques, but often also for significantly improved utilization of GPU computing compared to solvers based on the Cholesky decomposition.

Q^ⓔPyTorch provides (1) significant GPU acceleration (through MVM based inference); (2) state-of-the-art implementations of the latest algorithmic advances for scalability and flexibility (SKI/KISS-GP, stochastic Lanczos expansions, LOVE, SKIP, stochastic variational deep kernel learning, ...); (3) easy integration with deep learning frameworks.

Examples, Tutorials, and Documentation

See our documentation, examples, tutorials on how to construct all sorts of models in Q^ⓔPyTorch.

Installation

Requirements:

• Python >= 3.10
• PyTorch >= 2.2
• GPyTorch >= 1.13

Install Q^ⓔPyTorch using pip or conda:

pip install qpytorch
conda install qpytorch -c qpytorch

(To use packages globally but install Q^ⓔPyTorch as a user-only package, use pip install --user above.)

Latest (Unstable) Version

To upgrade to the latest (unstable) version, run

pip install --upgrade git+https://github.com/cornellius-gp/linear_operator.git
pip install --upgrade git+https://github.com/cornellius-gp/gpytorch.git
pip install --upgrade git+https://github.com/lanzithinking/qepytorch.git

Development version

If you are contributing a pull request, it is best to perform a manual installation:

git clone https://github.com/lanzithinking/qepytorch.git qpytorch
cd qpytorch
pip install -e .[dev,docs,examples,keops,pyro,test]  # keops and pyro are optional

Citing Us

If you use Q^ⓔPyTorch, please cite the following papers:

Li, Shuyi, Michael O'Connor, and Shiwei Lan. "Bayesian Learning via Q-Exponential Process." In Advances in Neural Information Processing Systems (2023).

@inproceedings{li2023QEP,
  title={Bayesian Learning via Q-Exponential Process},
  author={Li, Shuyi, Michael O'Connor, and Shiwei Lan},
  booktitle={Advances in Neural Information Processing Systems},
  year={2023}
}

Contributing

See the contributing guidelines CONTRIBUTING.md for information on submitting issues and pull requests.

The Team

Q^ⓔPyTorch is primarily maintained by:

• Shiwei Lan (Arizona State University)

We would like to thank our other contributors including (but not limited to) Shuyi Li, Guangting Yu, Zhi Chang, Chukwudi Paul Obite, Keyan Wu, and many more!

Acknowledgements

Development of Q^ⓔPyTorch is supported by.

License

Q^ⓔPyTorch is MIT licensed.