Operators are overloaded so that a normal looking function definition provides access to not only evaluations of itself, but to evaluations of any of its analytic derivatives.
Automatic differentiation means the user does not need to define the analytic expressions for all the various partial derivatives. It also means that those complex expressions are essentially calculated at compile time, and merely evaluated at runtime.
First order derivatives only, forward accumulation.
Choose "files" and either the "initial_submission" directory for the Ada version, or the other directory "cpp_AD" for the C++ version.
- Ada/C++ source code only. See files adaAD_*.tar.gz or cpp_AD_*.tar
- Examples are included that demonstrate a damped Newton's method for finding roots of systems of nonlinear equations.
- C++ Source Templates now available too, with equivalent capabilities. See file cpp_AD*.tar
Be the first to post a review of AdaAutoDiff!