Automatic differentiation in MATLAB (Q1184723)

The authors report on their implementation of automatic differentiation to generate the gradient of a multivariate real function. They define a C language function GRAD using the forward mode. GRAD can be called from MATLAB with a character string representing a function \(f\) of \(n\) variables and a point \(x\) in \(n\)-space at which \(f\) and its gradient are to be evaluated. A vector of \(1+n\) values is returned from GRAD representing \(f(x)\) and the gradient of \(f\) at \(x\). An example (solving a system of nonlinear equations with Newton's method) illustrates the use of GRAD.