Thank you all! On 14 September 2016 at 00:41, Steven G. Johnson <stevenj....@gmail.com> wrote:
> Both ForwardDiff and ReverseDiff source solve a different problem (taking > the derivative of a user-supplied function f(x)). The Matlab and NumPy > gradient functions, instead, take an array (not a function) and compute > differences of adjacent elements of the array, returning a new array. >