No label defined (Q3979896)

K. F. Gauss (1809) in ``Theoria motus corporum coelestium\dots'' has considered the problem of existence of the so-called pseudosolution of an overdetermined system of linear algebraic equations \(Ht=y\) where the vector \(y\) and some of columns of the matrix \(H\) are known approximately only (e.g. from an observation/experiment). Such a system may be written in the form \(Xt^{(1)}+At^{(2)}=y\), where \(X\) and \(A\) are \(n\times m\) and \(n\times p\) real matrices respectively, \(t^{(1)}\in\mathbb{R}^m\), \(t^{(2)}\in\mathbb{R}^p\), \(y\in\mathbb{R}^n\), \(m+p\leq n\). A similar notion of the `` quasisolution on a subset \(M\) of \(\mathbb{R}^{m+p}\) of such a system of equations was introduced by \textit{V. K. Ivanov} [Sov. Math., Dokl. 3, 981--983 (1962); translation from Dokl. Akad. Nauk SSSR 145, No. 2, 270--272 (1962; Zbl 0201.46701)]. The author defines some generalizations of these two notions and studies their existence, uniqueness and mutual relations (identities). He gives also constructions of these generalized ``solutions'' in some particular cases, namely when \(X\) and \(A\) are full rank matrices (i.e. \(rank[X,A]=m+p\)) or not and when alternatively the Euclidean norm or the energetic (weighted) one are used to measure the data approximation.