| Popis: |
We are interested in finding out the best way to smooth noisy data using cubic B-splines. Two small parameters are involved: the noise level ϵ, and the splines path h. The error in maximum norm between the actual data and the smoothed noisy ones is proved to be an O ( ϵ + h 2 ) , which leads to the choice of a splining path h = O ( ϵ ) . Error estimates on the derivatives are also provided, and a splining protocol is derived from them, which ensures an O ( ϵ 1 2 p ) maximum norm error on the p-th derivatives. To cite this article: S. Chaabane et al., C. R. Acad. Sci. Paris, Ser. I 346 (2008). |
