On updating triangular products of Householder reflections (Q801636)

Numerical algorithms for computing orthogonal factorizations often yield the orthogonal matrix as a product of n elementary reflections (Householder transformations). It is shown how such a product can be updated when the orthogonal matrix is multiplied by an elementary rotation or reflection. This update needs 5/2 \(n^ 2\) multiplications for a reflection and \(3n^ 2\) for a rotation, but preserves the compact storage used for the product form of the complete orthogonal matrix.