https://github.com/bgranzow/diff

automatic differentiation
https://github.com/bgranzow/diff

Last synced: 2 months ago
JSON representation

automatic differentiation

• Host: GitHub
• URL: https://github.com/bgranzow/diff
• Owner: bgranzow
• License: mit
• Created: 2015-08-10T17:30:41.000Z (over 9 years ago)
• Default Branch: master
• Last Pushed: 2015-08-11T20:35:17.000Z (over 9 years ago)
• Last Synced: 2024-06-11T17:01:10.087Z (7 months ago)
• Language: C++
• Size: 164 KB
• Stars: 6
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
## diff

a utility for automatic differentiation

implements a forward automatic differentiation variable (FAD) via operator overloading

### diff.h

* a single header implementation of a FAD variable

* conceptually simple

* uses temporary variables at each operation step

* temporary variables are sloooow

### expr.h

* an additional header that implements FAD with expression templates

* eliminates the need for temporary variables during each operation

* expression templates are fast (usually)

### useful information

* [wikipedia](https://en.wikipedia.org/wiki/Automatic_differentiation) has a good introduction to the subject of automatic differentiation

* [this](http://link.springer.com/article/10.1007%2Fs007910000048#page-1) paper by Aubert et al. discusses FAD with expression templates and an application to computational fluid dynamics

* [this](http://arxiv.org/abs/1205.3506) paper by Phipps and Pawlowski contains a good discussion about FAD with expression templates, some of its limitations, and the potential to improve its performance