|
Readme File | Software | Documentation
Abstract
If you need to calculate derivatives of complicated functions and find
yourself either taking finite differences or writing the derivatives
algebraically and then translating the expressions into source code,
you may want to consider using automatic differentiation (AD). AD
exploits the classic theorems of differential calculus to propagate
information about derivatives through arithmetic operations. In this
way, derivatives of a function can be calculated using the same
program that calculates the function itself. Because no
approximations are made, derivatives are calculated with machine
accuracy, avoiding the errors inherent in finite differences, an
especially important consideration when higher order derivatives are
required.
MXYZPTLK is a library of C++ classes -- or "objects" -- for
performing automatic differentiation. Originally written at Fermilab
in 1989, with a "User's Guide" provided in 1990, it has undergone
refinements and improvements over the last six years. It was
originally announced outside Fermilab in Automatic Differentiation of
Algorithms: Theory, Implementation, and Application (SIAM Press, 1991)
and has been used in a variety of contexts.
Send questions or comments to mxyzptlk_support@fnal.gov.
|
