Piecewise-approximation methods for multidimensional curves (Q1073283)

The authors propose an algorithm for piecewise polynomial approximation of a set \(\{y_ 1(t),...,y_ s(t)\}\) of given functions on the basis of a data \(\{y_ p(t_ j)\}\), \(p=1,...,s\), \(j=1,...,N\). Let \(F_ j^{(i)}\) be the polynomial of order m approximating \(y_ i\) on \([T_{j-1},T_ j]\) and \(\epsilon_ i(T_{j-1},T_ j)=\sum_{t_{\ell}\in (T_{j-1},T_ j]}[y_ i(t_{\ell})-F_ j^{(i)}]^ 2.\) A partition T of \([t_ i,t_ N]\) is constructed which minimizes a functional of the form \[ J(T)=\frac{1}{N}\sum^{s}_{i=1}\mu_ i\sum^{r}_{j=1}\frac{n_ j}{n_ j-m}\epsilon_ i(T_{j-1},T_ j) \] where \(\{\mu_ i\}\) are given weights, \(n_ j\) is the number of points \(\{t_ k\}\) in \((T_{j- 1},T_ j]\).