Zero-preserving iso-spectral flows based on parallel sums (Q864557)

Let \(\Delta\) be a set of pairs \((i,j)\) of integers between \(1\) and \(n\) containing also diagonal pairs \((i,i)\) and symmetric pairs \((i,j)\) and \((j,i)\). \(\Delta\) is regarded as a symmetric sparsity pattern of interest of nonzero entries for matrices. Let Sym\((n)\) denote the vector space of symmetric, \(n \times n\), real matrices and let Sym\((\Delta)\) denote the subspace of Sym\((n)\) consisting of the symmetric matrices which are zero outside the pattern \(\Delta\). In this paper, the problem of finding flows in the space Sym\((\Delta)\) which preserve eigenvalues and converge to diagonal matrices is considered. Earlier, \textit{K. R. Driessel} [Linear Algebra Appl.~379, 353-379 (2004; Zbl 1040.15012)] introduced the notion of quasi-projection onto the range of a linear transformation from one inner product space to another. Here, the notion of quasi-projection onto the intersection of the ranges of two linear transformations \(A\) and \(B\) from two inner product spaces into a third inner product space is introduced. The quasi-projection method for designing a new flow corresponding to the above mentioned problem is described. Especially, an implementation of the iso-spectral flows on the space of symmetric zero-preserving matrices in Matlab is described. Several numerical experiments performed with this computer program are presented. In all these computations this flow converges to a diagonal matrix and thus these computations give evidence for this conjecture. Conditions when the zero-preserving flow coincides with the Toda flow are given.