A project implementing sparse spectrum Gaussian process regression using PyTorch and quasi-random Fourier features generated using Halton sequences.
- PyTorch >= 1.1 (https://pytorch.org/)
- ghalton >= 0.6.1 (https://github.com/fmder/ghalton)
- 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
- Clone git repository
git clone git@bitbucket.org:rafaol/ssgp.git
- Switch to repository's root directory:
cd ssgp/
- Install dependencies:
pip install -r requirements.txt
- Install package in editable mode (-e option) to keep track of updates:
pip install -e .