Vector measure orthonormal systems and self-weighted functions approximation (Q2580238)

In this paper, \(\lambda\) is a positive \(\ell_{2}\)-valued measure on a measurable space \(( \Omega, \Sigma)\); \( \| .\| _{\lambda}\) will denote the norm in \( L_{2}(\lambda) \). Though in general \( L_{2}(\lambda) \) is not a Hilbert space, some Hilbert space type properties of its subspaces are proved. A countable set of functions \( \mathcal{C} = \{ h_{n}: n \in \mathbb N \} \subset L_{2}(\lambda)\) is called \(\lambda\)-orthonormal if (i) \(\| f_{n} \| _{\lambda} =1\) for all \(n\), (ii) \( \int f_{n} f_{m} \,d \lambda =0 \) for all \(n\) and \(m\) with \(n\neq m\), and (iii) \(T_{\mathcal{C}}\to L_{2}(\lambda): \ell_{2} (\ell_{2})\), \(T_{\mathcal{C}}(f)= \{ \int f h_{i} d \lambda \}\) is continuous. \( \mathcal{S} (\mathcal{C})\) denotes the closed subspace generated by \( \mathcal{C}\) and \( \mathcal{S}^{\lambda} (\mathcal{C}) = \{ f \in L_{2}(\lambda): \int f g d \lambda=0\) for every \( g \in \mathcal{S} (\mathcal{C})\}\). For \( f, g\) in \( L_{2}(\lambda) \), the \(\mathcal{C}\)-distance between them is defined as \(d_{\mathcal{C}} (f,g)= \| T_{\mathcal{C}}(f-g) \| _{2,2}\); the \(\mathcal{C}\)-distance between \(f\) and a subspace \(S \subset L_{2}(\lambda) \) is \(d_{\mathcal{C}} (f,S) = \inf_{g \in S} \| T_{\mathcal{C}}(f-g) \| _{2,2}\). The following are the main results: \( \mathcal{S} (\mathcal{C})\) is isomorphic to \(\ell_{2}\) and every \( g \in \mathcal{S} (\mathcal{C})\) can be written in the form \( g= \sum_{i=1}^{\infty} \alpha_{i} h_{i}\) with \( \{ \alpha_{i} \} \in \ell_{2}\). There is a continuous projection \(P_{\mathcal{C}}: L_{2}(\lambda) \to \mathcal{S} (\mathcal{C})\) such that, for a \(g \in L_{2}(\lambda) \), the \(\mathcal{C}\)-orthogonal distance between \(g\) and \( \mathcal{S} (\mathcal{C})\) is attained by the function \(P_{\mathcal{C}}(g)= \sum_{i=1}^{\infty} \beta_{i} h_{i}\), where \( \beta_{i} = \langle \int g h_{i} \,d \lambda, \int h_{i}^{2} \,d \lambda \rangle\). Also, for \(g \in L_{2}(\lambda) \), \( g - P_{\mathcal{C}}(g) \in \mathcal{S}^{\lambda} (\mathcal{C})\) iff \( \| T_{\mathcal{C}}(g- P_{\mathcal{C}}(g)) \| _{2,2}=0\); in this case, \( P_{\mathcal{C}}(g)\) is the function of \(\mathcal{S} (\mathcal{C})\) for which \( \inf_{f \in \mathcal{S} (\mathcal{C})} \| g - f \| _{\lambda}\) is attained. Finally, some geometrical descriptions and interpretations of these results are given.