Linear preserving \(gd\)-majorization functions from \(M_{n,m}\) to \(M_{n,k}\) (Q648486)

A real \(n\)-by-\(m\) matrix \(A\in M_{n\times m}({\mathbb R})\) is directionally g-majorized by \(B\in M_{n\times m}({\mathbb R})\) if for each vector \(x\) there exists a g-doubly stochastic matrix \(G\) (i.e., row and column sums of \(G\) all equal to one) with \(Ax=GBx\), and \(A\) is multivariately g-majorized by \(B\) if \(G\) does not depend on \(x\). The authors prove that directional and multivariate g-majorization coincide on \(M_{n\times m}({\mathbb R})\). It is also shown that linear maps \(T:M_{n\times m}({\mathbb R})\to M_{n\times k}({\mathbb R})\) which preserve directional g-majorization come in one of three different types, which are not given in this short review, and only the second of the these three types can preserve directional g-majorization in both directions.