There are a variety of additional numerical indicators which can be used to characterize sparse matrices. For example: extreme eigenvalues, Frobenius norm, maximum element, percentage of weakly diagonally dominant rows, columns, average number of nonzeros per row, column, number of elements in (shortest, longest) (row, column), upper and lower bandwidth, average bandwidth, bandwidth of closest 95%, 90%, 80% to diagonal, standard deviation of distance of elements to diagonal, number of nonzero diagonals, and percent of symmetric elements. Most of these, and others, are computed by tools such as the INFOFUN routine of SPARSKIT (Saad, 1994). Since such data might only be of limited value in searching we have not included it in our database. However, we do intend to run a standard code such as INFOFUN against all the matrices in the collection, and include such data in the matrix home pages.

Thu May 30 12:55:31 EDT 1996