This package implements higher dimensional function approximation by setting up interpolation matrices using different combination methods, including the sparse grid method. After the setup, the approximation problem can be solved by either solving the least squares problem or by using any gradient descent method. The package builds on top of the PyTorch package.
The package can be downloaded from Github:
git clone https://github.com/timotheehornek/sparsetorch.git
Afterwards, it can be installed using pip. Simply run the following after cloning:
pip install ./path_to_folder/sparsetorch/
Note that torch==1.9.0+cpu (for the package) and matplotlib==3.4.3 (for the examples) are also required.
PyTorch can be downloaded here.
Documentation can be found in the docs folder and additionally hosted here.
A tutorial is provided as Jupyter notebook. Moreover, several examples are provided for demonstrating the capabilities in the same folder.