On the uniqueness of generalized monosplines of least \(L_ p\)-norm (Q1082552)

The unique solvability of a minimization problem from spline theory is investigated. Let \(K: [t^-,t^+]\times [x^-,x^+] \to {\mathbb{R}}\) be an extended totally positive kernel in the sense of \textit{S. Karlin} [Total positivity (1968; Zbl 0219.47030)]. Moreover, let a function \(f: [t^-,t^+]\to {\mathbb{R}}\) be defined by \[ f(t)=\int^{x^+}_{x^- }K(t,x) sgn \prod^{n}_{i=1}(x_ i-x)^ m\quad d\mu (x)\quad (t\in [t^-,t^+]), \] where \(x_ 1,...,x_ n\in [x^-,x^+]\) and \(d\mu\) is a nonnegative measure. A function \(M:[t^-,t^+]\to {\mathbb{R}}\), defined by \[ M(t)=f(t)-\sum^{n}_{i=1}\sum^{m- 2}_{j=0}a_{ij}K_ j(t,x_ i)\quad (t\in [t^-,t^+]), \] where \(a_{ij}\in {\mathbb{R}}\) and \(K_ j(t,x)=(\partial /\partial x)^ jK(t,x),\) is called generalized monospline. For \(m=2\) this function is the usual monospline and for \(m=1\) (where \(\sum^{-1}_{i=1}=0)\) it is the perfect spline. The authors study the problem, if the parameters \(x_ i\) and \(a_{ij}\) which minimize the \(L_ p\)-norm \((1<p\leq \infty)\) of M are uniquely determined. It is proved that if \(K(t,x)=g(t-x)\) and \(d\mu =dx\), then there exists a unique generalized monospline of least \(L_ p\)-norm \((1<p<\infty)\). Moreover, the authors show that if \(d\mu (x)=w(x)dx\), where \(w: [x^- ,x^+]\to {\mathbb{R}}\) is a continuous nonnegative weight function, then there exists a unique generalized monospline of least \(L_{\infty}\)-norm. These theorems are obtained by applying degree theory and extend known results on usual monosplines and perfect splines.