\(r\)-numbers completion problems of partial upper canonical form. II (Q1861813)

In part I of this paper [ibid. 15, 685-691 (2002; Zbl 1006.65039)], the authors solved the \(r\)-number completion problem for the pair \((A,B)\) which completes a partially specified \(n\times n\) matrix \(A\), given the \(n\times m\) matrix \(B\) which has only \(r_1\) nonzero linearly independent rows. The completion is such that the \(r\)-numbers of the pair \((A,B)\) are equal to a pre-specified set of values \(r_1,\ldots,r_p\). The problem was solved by giving algorithms that worked under some restrictive conditions. For example, it was required that the nonzero columns in \(B\) had indices \(i_k\), \(k=1,\ldots,r_1\) with \(i_1<i_2<\cdots<i_{r_1}\leq \lfloor (n+1)/2\rfloor\). In this paper, the case \(i_1,\ldots,i_{r_1}\) with \(i_1<i_2<\cdots<i_{r_1}\geq \lfloor (n+1)/2\rfloor\) is considered. Also for the general case with no restrictions on the indices \(i_k\) an algorithm is formulated. It is conjectured that it constructs a completion whose \(r\)-numbers majorize the sequence of \(r\)-numbers of any other completion.