An Algorithm for Computing Derivatives of Noisy Data
ARMY BALLISTIC RESEARCH LAB ABERDEEN PROVING GROUND MD
Pagination or Media Count:
An algorithm for computing derivatives of noisy data, as described by the authors in two previous reports, is modified, and empirical support for the choice of relative weights for approximating functions is provided. The algorithm is applied to ten sets of synthetic data generated by evaluating analytic functions and addition pseudo-random errors to their values. The accuracy of the estimates of the first and second derivatives is on the average almost five times better than that obtained by cubic B-spline approximation. Extrapolation is considered.
- Theoretical Mathematics