https://github.com/john-science/expm

Trying to improve performance on Matrix Exponentials in Python
https://github.com/john-science/expm

expm matricies matrix-exponentials python

Last synced: about 2 months ago
JSON representation

Trying to improve performance on Matrix Exponentials in Python

• Host: GitHub
• URL: https://github.com/john-science/expm
• Owner: john-science
• License: gpl-3.0
• Created: 2020-04-12T13:10:50.000Z (over 4 years ago)
• Default Branch: main
• Last Pushed: 2024-04-22T22:36:45.000Z (9 months ago)
• Last Synced: 2024-11-08T19:04:49.132Z (2 months ago)
• Topics: expm, matricies, matrix-exponentials, python
• Language: Python
• Homepage:
• Size: 55.7 KB
• Stars: 1
• Watchers: 1
• Forks: 0
• Open Issues: 1
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# expm

[![Build Status](https://travis-ci.com/john-science/expm.svg?branch=master)](https://travis-ci.com/john-science/expm)

[![codecov](https://codecov.io/gh/john-science/expm/branch/master/graph/badge.svg)](https://codecov.io/gh/john-science/expm)

> Improving the performance of Matrix Exponentials in Python

This project exists purely out of necessity. The 'truth benchmark' I will use for all testing here will be the SciPy implementation of matrix exponentials

([scipy.linalg.expm](https://docs.scipy.org/doc/scipy/reference/generated/scipy.linalg.expm.html)).

I would like to point out that the fastest implementation I have found in preparation for this project was from [rngantner on GitHub](https://github.com/rngantner/Pade_PyCpp/blob/master/src/expm.py).

The first step in trying to make performant code is to set strict bounds on the problem you are willing to solve. So, the restrictions I will put on my implentation of `expm` will be:

* The code must be callable from Python v3.5 to v3.7.

* The performance testing will be done on Python v3.6.

* The matrix will be real-valued.

* The performance will only be tested against 2D arrays.

* The performance will not be tested on matricies of size < 10x10.

* The performance will not be tested on matricies of size > 100x100.