On approximation of multivalued mapping by algebraic polynomial with constraints (Q2356571)

Let \(n\) and \(N\) be integers such that \(n\geq 0\) and \(N\geq n+1\). \(p_n\left(A,t\right)\) denotes the algebraic polynomial with coefficient vector \(A=\left(a_0,a_1,\dots,a_n\right)\in{\mathbb R}^{n+1}\), i.e., \(p_n\left(A,t\right)=a_0+a_1 t+\dots+a_n t^n\). On the set \(T=\left\{t_0<t_1<\dots<t_N\right\}\), the multivalued mapping function \(\Phi\) is defined by \(\Phi\left(t_k\right)=\left[y_{1,k};y_{2,k}\right]\), \(y_{1,k}\leq y_{2,k}\) (\(k=0,1,\dots,N\)). It is also assumed that we are given \(\nu_{1,k}\), \(\nu_{2,k}\) with \(\nu_{1,k}<\nu_{2,k}\), where \(k\in S:=\left\{s_1<\dots<s_m\right\}\subset\left\{0,1,\dots,N\right\}\), \(m\geq 1\). Let us denote \[ f\left(A,t_k\right)=\max\left\{p_n\left(A,t_k\right)-y_{1,k},y_{2,k}-p_n\left(A,t_k\right)\right\},\quad k=0,1,\dots,N, \] \[ g\left(A,t_k\right)=\max\left\{\nu_{1,k}-p_n\left(A,t_k\right),p_n\left(A,t_k\right)-\nu_{2,k}\right\},\quad k\in S, \] \[ \gamma\left(A\right)=\max_{k\in S}g\left(A,t_k\right). \] The article is devoted to the following two problems: { Problem~1:} \[ \rho\left(A\right):=\max_{k=0,1,\dots,N}f\left(A,t_k\right)\to\min_{A\in D=\left\{A\in{\mathbb R}^{n+1}:\,\gamma\left(A\right)\leq 0\right\}}; \] {Problem~2:} \[ \rho\left(A\right):=\max_{k=0,1,\dots,N}f\left(A,t_k\right)\to\min_{A\in{\mathbb R}^{n+1}}. \] The author found sufficient and necessary conditions for a vector \(A^*\in{\mathbb R}^{n+1}\) to be a solution to Problem~1 or to Problem~2. Some examples are also considered. The article should be interesting for specialists in approximation theory, numerical analysis and other areas, where the best polynomial approximation is engaged.