Euclidean distances and least squares problems for a given set of vectors (Q2382757)

Given \(n\) real vectors of size \(m\), \(n\) simple least squares problems can be formed in the following fashion. Fix one of the vectors and consider the remaining \(n-1\) vectors as the coefficients of the matrix problem. Let the original vectors be arranged on a matrix, \(A\), of size \(m\times n\). The authors of this paper show a simple matrix identity relating \(A\) and the 2-norm of each square residual. As an application of this result, they consider an improvement upon an algorithm in the theory of communication, where the QR factorisation of a certain column permutation of \(A\) is computed.