Exact Markov-type inequalities for oscillating perfect splines (Q5956458): Difference between revisions

In this interesting and elegant paper, the authors establish Markov inequalities with sharp constants for oscillating perfect splines. The result is a corollary of a general ``inheritance principle'' proved by the authors: certain inequalities involving polynomials or oscillating perfect splines, are inherited by their derivatives. Recall that a perfect spline with knots \(\xi_1< \xi_2<\cdots <\xi_{n-r}\) has the form \[ s(t)= \sum^r_{j=1} \alpha_j t^{j-1}+ c\left( t^r+2 \sum^{n-r}_{i=1}(-1)^i (t-\xi_i)^r_+\right), \] where \(c,\alpha_1, \alpha_2,\dots, \alpha_r\) are real numbers. Let \({\mathcal P}_{nr}\) denote the set of all such perfect splines, which have exactly \(n\) simple zeros in \([-1,1]\). An important perfect spline is the Chebyshev perfect spline \(T_{nr}\), the unique spline of the above form, that takes its maximal value of 1 on \([-1,1]\), with alternating sign, at \(n+1\) points in \([-1,1]\). The authors use their inheritance principle to show that for \(s\in{\mathcal P}_{nr}\), \(1\leq p\leq \infty\) and \(1\leq k\leq r\), \[ \|s^{(k)}\|_{L_p[-1,1]} \leq\|T_{nr}^{(k)} \|_{L_p[-1,1]} \|s\|_{C[-1,1]}. \] They also establish a similar inequality in a more general setting, where the \(L_p\) norm is replaced by a convex function of \(|s|\). The proofs display a deep understanding of classical ideas of equioscillation, interlacing, delicate zero counting, and spline theory. It is essential reading for researchers in approximation theory.