next up previous
Next: CSI-MSVD Performance Results Up: Computational Environments Previous: Performance Evaluation of

4.3 Test Problems and Parameters

  The test problems used to evaluate CSI-MSVD were obtained primarily from information retrieval applications and from seismic reflection tomography problems (see [3,4,22]). Table 1 lists the sizes and sparsity of the matrices used for evaluating the performance of the CSI-MSVD algorithm. Here sparsity is defined by

where nnz is the number of non-zeros, nrows is the number of rows, and ncols is the number of columns of the sparse matrix. In order to study the effect of clustered spectra on the accuracy of the error estimate, synthetic diagonal test matrices having clustered or multiple diagonal elements were also used as input (see [22]).

Table 1: Test matrices used to evaluate the performance of CSI-MSVD.

Michael W. Berry (
Sun May 19 11:34:27 EDT 1996