| |
| | [No title] |
 | | Also states the % `fundamental transformation formula for divided differences', something % that reduces to the formula, % \sum_{j=1}^n \psi_{1,j-1}\dvd{x_1,\ldots,x_j} + % \sum_{j>n} \psi+_{j-n+1,j-1}(x_j-x_{j-n})\dvd{x_{j-n},\ldots,x_j} % with \psi^+_{r,s}:= (\cdot-x_r)_+\cdots(\cdot-x_s)_+, for the interpolant % at the sequence x_1< x_2< \cdots that agrees, on [x_j\fromto x_{j+1}], with % the polynomial interpolant at x_{j-n+1},\ldots,x_j. |
 | | Evaluation of this % formula at x_i provides a unique description of the linear functional [x_i] % in terms of the linear functionals \dvd{x_{max(1,j-n+1)},\ldots,x_{j-1}}, % j=1,2,\ldots, hence Popoviciu's name `transformation formula' for the % latter. |
 | | Also proves the `mean-value formula for divided differences': % \dvd{t_0,\ldots,t_n} = \sum_j a_j(t,s)\dvd{s_j,\ldots,s_{j+n}} % for any monotone refinement s of monotone t, with a_j(t,s)\ge0 and summing % to 1. |
| www.cs.wisc.edu /~deboor/bib/P (405 words) |
|