detkit Documentation#

A python package to compute common functions involving determinant of matrices used in machine learning.

GitHub

https://github.com/ameli/detkit

PyPI

https://pypi.org/project/detkit/

Anaconda Cloud

https://anaconda.org/s-ameli/detkit

Docker Hub

https://hub.docker.com/r/sameli/detkit

Supported Platforms#

Successful installation and tests have been performed on the following platforms and Python/PyPy versions shown in the table below.

Platform	Arch	Python Version				PyPy Version ¹			Continuous Integration
Platform	Arch	3.9	3.10	3.11	3.12	3.8	3.9	3.10	Continuous Integration
Linux	X86-64	✔	✔	✔	✔	✔	✔	✔
Linux	AARCH-64	✔	✔	✔	✔	✔	✔	✔
macOS	X86-64	✔	✔	✔	✔	✔	✔	✔
macOS	ARM-64	✔	✔	✔	✔	✔	✔	✔
Windows	X86-64	✔	✔	✔	✔	✔	✔	✔
Windows	ARM-64 ²	✔	✔	✔	✔	✔	✔	✔

Python wheels for detkit for all supported platforms and versions in the above are available through PyPI and Anaconda Cloud. If you need detkit on other platforms, architectures, and Python or PyPy versions, raise an issue on GitHub and we build its Python Wheel for you.

^{1. Wheels for PyPy are exclusively available for installation through pip and cannot be installed using conda.}

^{2. Wheels for Windows on ARM-64 architecture are exclusively available for installation through pip and cannot be installed using conda.}

Install#

Install with pip from PyPI:

pip install detkit

Install with conda from Anaconda Cloud:

conda install -c s-ameli detkit

For complete installation guide, see:

Install

Docker#

The docker image comes with a pre-installed detkit, an NVIDIA graphic driver, and a compatible version of CUDA Toolkit libraries.

Pull docker image from Docker Hub:

docker pull sameli/detkit

For a complete guide, see:

Docker

List of Functions#

`detkit.logdet`	Compute the logdet of a matrix.
`detkit.loggdet`	Compute the loggdet of a matrix.
`detkit.logpdet`	Compute the logpdet of a matrix.
`detkit.memdet`	Compute log-determinant under memory constraint.

See API Reference for the full list of functions.

Tutorials#

Launch an online interactive tutorial in Jupyter notebook.

Benchmarks#

See benchmark test for evaluating the numerical performance of the functions in real applications.

Features#

Functions are implemented with a novel algorithm described in [1].
The underlying library is implemented in C++ and wrapped in cython.
An accurate count of computational FLOPs during the execution of functions can be measured.

How to Contribute#

We welcome contributions via Github’s pull request. If you do not feel comfortable modifying the code, we also welcome feature request and bug report as Github issues.

How to Cite#

If you publish work that uses detkit, please consider citing the following manuscripts.

[1]

Ameli, S., and Shadden. S. C. (2023). A Singular Woodbury and Pseudo-Determinant Matrix Identities and Application to Gaussian Process Regression. Applied Mathematics and Computation 452, 128032. DOI

@article{amc-2023,
    title = {A singular Woodbury and pseudo-determinant matrix identities and application to Gaussian process regression},
    journal = {Applied Mathematics and Computation},
    volume = {452},
    pages = {128032},
    year = {2023},
    issn = {0096-3003},
    doi = {https://doi.org/10.1016/j.amc.2023.128032},
    author = {Siavash Ameli and Shawn C. Shadden},
}

[2]

Siavash Ameli, Chris van der Heide, Liam Hodgkinson, Fred Roosta, Michael W. Mahoney (2024). Determinant Estimation under Memory Constraints and Neural Scaling Laws (under review) DOI

@article{nip-2024,
    title = {Determinant Estimation under Memory Constraints and Neural Scaling Laws},
    journal = {},
    volume = {},
    pages = {},
    year = {},
    issn = {},
    doi = {},
    author = {Siavash Ameli, Chris van der Heide, Liam Hodgkinson, Fred Roosta, Michael W. Mahoney},
}

[3]

Ameli, S. (2022). DetKit, a python package for matrix determinant.

@misc{zenodo.6395320,
    title = {{DetKit}, a python package for matrix determinant},
    author = {Ameli, S.},
    year = {2022},
    howpublished = {\url{https://pypi.org/project/detkit/}}
}