AQZ-method based on semiseparable matrices (Q932737)

The authors aim at implementing an alternative method for solving the generalized eigenvalue problem, based on structured rank matrices. This new approach involves the reduction of a pair of two general matrices \((A,B)\) to a lower semiseparable, upper triangular matrix pair. An algorithm is presented to accomplish this reduction, together with the associated \(QZ\)-iteration. Some numerical examples show that the accuracy of this algorithm is comparable to the accuracy of the \texttt{eig}-routine of MATLAB\(^{\circledR}\).