wayneouyang / gpflow

Gaussian processes in TensorFlow

GPflow

GPflow is a package for building Gaussian process models in python, using TensorFlow. It was originally created and is now managed by James Hensman and Alexander G. de G. Matthews. The full list of contributors (in alphabetical order) is Artem Artemev, Rasmus Bonnevie, Alexis Boukouvalas, Ivo Couckuyt, Keisuke Fujii, Zoubin Ghahramani, David J. Harris, James Hensman, Pablo Leon-Villagra, Daniel Marthaler, Alexander G. de G. Matthews, Tom Nickson, Valentine Svensson, Mark van der Wilk. GPflow is an open source project so if you feel you have some relevant skills and are interested in contributing then please do contact us.

What does GPflow do?

GPflow implements modern Gaussian process inference for composable kernels and likelihoods. The online user manual contains more details. The interface follows on from GPy, for more discussion of the comparison see this page.

GPflow uses TensorFlow for running computations, which allows fast execution on GPUs, and uses Python 3.5 or above.

Install

1) Quick install

GPflow can be installed by cloning the repository and running

pip install .

in the root folder. This also installs required dependencies including TensorFlow. When GPU support is needed, a manual installation of TensorFlow is recommended (next section), as one cannot rely on pip to get this running.

2) Alternative method

A different option to install GPflow requires installation of TensorFlow first. Please see instructions on the main TensorFlow webpage. You will need at least version 1.3 (we aim to support the latest version). We find that for most users pip installation is the fastest way to get going. Then, for those interested in modifying the source of GPflow, we recommend

python setup.py develop

but installation should work well too:

python setup.py install

You can run the tests with python setup.py test.

We document the version history.

Docker image

We also provide a Docker image which can be run using

docker run -it -p 8888:8888 gpflow/gpflow

The image can be generated using our Dockerfile.

Getting help

Please use GitHub issues to start discussion on the use of GPflow. Tagging enquiries discussion helps us distinguish them from bugs.

Contributing

All constructive input is gratefully received. For more information, see the notes for contributors.

Compatibility

GPflow heavily depends on TensorFlow and as far as TensorFlow supports forward compatibility, GPflow should as well. The version of GPflow can give you a hint about backward compatibility. If the major version has changed then you need to check the release notes to find out how the API has been changed.

Unfortunately, there is no such thing as backward compatibility for GPflow models, which means that a model implementation can change without changing interfaces. In other words, the TensorFlow graph can be different for the same models from different versions of GPflow.

Projects using GPflow

A few projects building on GPflow and demonstrating its usage are listed below.

Project	Description
GPflowOpt	Bayesian Optimization using GPflow.
VFF	Variational Fourier Features for Gaussian Processes.
Doubly-Stochastic-DGP	Deep Gaussian Processes with Doubly Stochastic Variational Inference.
BranchedGP	Gaussian processes with branching kernels.
gpflow-monitor	Tools for monitoring and checkpoining optimisation.

Let us know if you would like your project listed here.

Citing GPflow

To cite GPflow, please reference the JMLR paper. Sample Bibtex is given below:

@ARTICLE{GPflow2017,
   author = {Matthews, Alexander G. de G. and {van der Wilk}, Mark and Nickson, Tom and
	Fujii, Keisuke. and {Boukouvalas}, Alexis and {Le{\'o}n-Villagr{\'a}}, Pablo and
	Ghahramani, Zoubin and Hensman, James},
    title = "{{GP}flow: A {G}aussian process library using {T}ensor{F}low}",
  journal = {Journal of Machine Learning Research},
  year    = {2017},
  month = {apr},
  volume  = {18},
  number  = {40},
  pages   = {1-6},
  url     = {http://jmlr.org/papers/v18/16-537.html}
}