Differentiation · CMNA Laboratory
Move a finite distance away from x, measure the change in f, and turn the resulting secant slope into an approximation of the tangent.
A large h measures the wrong neighborhood. A very small h eventually runs into floating-point subtraction. The interesting numerical question is not simply “make h smaller,” but how the approximation behaves as h changes.
Bring your own function
R evaluates your function while the laboratory halves h repeatedly and compares forward, symmetric, and Richardson estimates.
Preparing R…f and optionally fp| level | h | forward | symmetric | Richardson |
|---|