What is automatic differentiation?
Automatic differentiation (AD) is a method to evaluate the derivative of a function. Unlike, symbolic differentiation, AD can work on extremely complex problems. And AD methods are more precise than the approximations that come from numerical differentiation. These calculations are done behind the scenes, so the user needs only to provide the code to calculate an objective function to be minimized. One minor challenge of using AD methods is that care needs to be taken that the objective function to be minimized is differentiable with respect to the parameters. To meet this challenge, ADMB provides some the smoothed alternatives to functions like absolute value that can be used to keep things differentiable. For more details, see Automatic Differentiation on Wikipedia.
