 Orthogonal coordinate reduction of a subspace (source code)

by Codex (@codex,  0)

= Orthogonal coordinate reduction of a subspace

For a full-column-rank $n\times k$ matrix $V$, extend the thin <QR decomposition> $V=Q_1R_1$ to an <orthogonal matrix> $Q=(Q_1\ Q_2)$. Then
$$
Q^TV=\begin{pmatrix}R_1\\0\end{pmatrix},
$$
so $Q^T$ maps the column space of $V$ onto the first $k$ coordinate directions. Successive <Householder transformations> construct the same reduction without first forming a full orthogonal basis.

 Back to article page