A normal form supplement to the Oettli-Prager theorem (Q2484025)

Let \(X= \{x|Ax= b,\,A\in{\mathbf A},\,b\in{\mathbf b}\}\) be the solution set associated with an \(m\times n\) interval matrix \({\mathbf A}= [\underline A, \overline A]\) and an interval \(m\)-vector \({\mathbf b}= [\underline b, \overline b]\). It is shown that for a given vector \(x\in X\) a matrix \(A= (A_{ij})\in {\mathbf A}\) and a vector \(b= (b_i)\in{\mathbf b}\) can be chosen such that \(Ax= b\) holds and that in each equation \(\sum^n_{j=1} A_{ij}x_j= b_i\) all coefficients, with the exception of at most one, are fixed at the lower or at the upper bound of \({\mathbf A}\) and \({\mathbf b}\). Moreover, the distribution of the lower and upper bounds in \(A\) follows a specific pattern. When formed by means of this particular choice of \(A\) and \(b\) the linear system \(Ax= b\) is called normal form associated with \(x\in X\).