Automatic Differentiation
According to Griewank's standard reference (Evaluating Derivatives- Principles & Techniques of Algorithmic Differentiation, SIAM 2000)
Algorithmic, or automatic, differentiation (AD) is concerned with the accurate and efficient evaluation of derivatives for functions defined by computer programs. No truncation errors are incurred, and the resulting numerical derivative values can be used for all scientific computations that are based on linear, quadratic, or even higher order approximations to nonlinear scalar or vector functions.
There is now worldwide interest in this exciting area, with applications to many areas of science and engineering. For further general information visit the international AD web-site
The AD Research Group within AMOR conducts internationally significant research in AD theory, tool development and applications. The AD Group comprises:
• Dr Shaun Forth
• Dr Mohamed Tadjouddine
• Mr Rahul Kharche
together with our colleague Dr John Pryce (CISE Group, Informatics & Simulation) and our visiting Professor of Numerical Analysis, John Reid.
Below you will find links to descriptions of our work in this area.
• European AD Workshops: twice yearly informal research meetings with location rotating around four leading European AD research groups.
• UK AD Workshops: archives of the former twice yearly informal research meetings
• EliAD Tool : World's first source-transformation implementation of the vertex elimination AD algorithm for Jacobian evaluation.
• Matlab Automatic Differentiation (MAD) Project: MAD features an efficient implementation of forward mode AD via operator overloading for functions defined by Matlab code.
• AD Enabled ODE Solvers Project: Joint work with Prof. Shampine (Southern Methodist University) investigating use of AD in MATLAB's ODE solvers to improve robustness, performance and ease of use.
• AD2CompEng: This EPSRC funded research project aims to apply present generation AD tools to significant Computational Engineering problems.
• Industrial Applications: We continue to collaborate with the UK Aerospace industry on differentiation of industrial strength CFD solvers.