1. What sort of error frustrates attempts to numerically differentiate a given smooth function to arbitrary accuracy? What is the step size h with smallest overall error?
2. Why is the numerical differentiation of noisy data a risky affair?
3. What is a differentiation matrix?