Generalized inverses under the \(\ell _ 1\)-norm (Q1092973)

For a system \(Ax=b\), a minimum generalized inverse (m.g.i.) G of A, in a normed space, satisfies \(\| Gy\| =\inf \| x\|\) where inf is taken over all solutions of \(Ax=y\). An approximate generalized inverse (a.g.i.) is a solution of \(AGA=A\) such that \(\| AGy-y\| =\min \| Ax-y\|\) for all y, where min is taken over x. This paper gives several conditions for existence, non-existence, and uniqueness of m.g.i. and a.g.i. for \(\ell_ 1\) spaces.