ssgp

A project implementing sparse spectrum Gaussian process regression using PyTorch and quasi-random Fourier features generated using Halton sequences.

Requirements

• PyTorch >= 1.1 (https://pytorch.org/)
• ghalton >= 0.6.1 (https://github.com/fmder/ghalton)

Optional

• NLopt (https://pypi.org/project/nlopt/): for hyper-parameters tuning
• Jupyter Notebook (https://jupyter.org/): to run example notebook
• Matplotlib (https://matplotlib.org/): to plot in example notebook

Installation

1. Clone git repository

git clone git@bitbucket.org:rafaol/ssgp.git

2. Switch to repository's root directory:

cd ssgp/

3. Install dependencies:

pip install -r requirements.txt

4. Install package in editable mode (-e option) to keep track of updates:

pip install -e .