Approximation and interpolation by convexity-preserving rational splines (Q1330859): Difference between revisions

The authors introduce the following rational splines: \[ R_ j(x)= c_{ij}(1- t)+ {c_{3j}(1- t)^ 3\over \lambda_ j t+1}+ c_{2j} t+ {c_{4j} t^ 3\over \lambda_ j(1- t)+ 1},\tag{1} \] where \(t= (x- x_ j)/\Delta_ i\) with \(\Delta_ j= x_{i_{j+1}}- x_{i_ j}\) and given data points with \(x_ 0< x_ 1<\cdots< x_ N\), \(R\in C^ 2[x_ 0, x_ N]\) the formula (1) gives the form of \(R\) on each knot interval \([x_{i_ j},x_{i_{j+1}}]\). The authors prove for a fixed parameter \(\alpha> 0\) that the optimal rational spline \(R^*_ \alpha\) exists technically. Moreover, the algorithm is presented in this paper for computing some kind of convexity-preserving rational spline, and it is associated with an interesting now nonlinear approximation theory problem -- a nonlinear constrained optimization problem. Then, four examples are showed.