cics-nd / gptorch

Gaussian processes with PyTorch

gptorch

Gaussian processes with PyTorch

Installation

Via pip

$ pip install gptorch

From source

You can now do:

$ pip install -r -requirements.txt
$ python setup.py install

Otherwise, we've provided an environment.yml to make a new virtual environment with Anaconda:

$ conda env create -f environment.yml
$ source activate gptorch

Models implemented:

• Vanilla GP regression model
• Sparse GP (variational free energy model)
• Sparse Variational GP

To be implemented:

• FITC sparse GPs
• Structured ("Kronecker") GPs
• Bayesian GPLVM
• Dynamical GP-LVM/Bayesian warped GP
• Non-Gaussian likelihoods (e.g. for classification)
• Correlated outputs
• Deep GPs