The periodic QR algorithm is a disguised QR algorithm (Q2502889): Difference between revisions

The author studies the eigenvalue computation of a matrix \(A=\mathbb{R}^{pn \times pn}\), of the form: \[ \left [ \begin{matrix} 0 & \cdots & \cdots & A^{(p)}\\ A^{(1)} & \cdots & 0 & 0\\ \cdots & \cdots & \cdots & 0\\ 0 & \cdots & A^{(p-1)} & 0 \end{matrix} \right ], \] where \(A^{(k)}\in \mathbb{R}^{n \times n},\;1 \leq k \leq p.\) He proves that the QR algorithm applied to a shuffled version of \(A\) preserves its structure and requires only \(O(pn^3)\) floating point operations with a preliminary step in the QR algorithm reducing a general matrix to its Hessenberg form.