Automatic numerical differentiation by discrete mollification (Q580880)

The author gives a method to estimate the derivative based on attempting to reconstruct a modified version of the derivative. This approach was first introduced by \textit{V. V. Vasin} [Zh. Vychisl. Mat. Mat. Fiz. 13, 1383-1389 (1973; Zbl 0306.65007)]. In this paper, the approximation is generated initially by filtering the noisy data by discrete convolution with a suitable averaging kernel and the use of centered differences to solve numerically the associated well-posed problem. Several numerical examples are also analyzed.