Skip to content

4.4-final-2006

Final Exam, 2006 · Section 4.4: Orthonormal Bases and Gram-Schmidt Source: finalexam_1.pdf, p. 147

Problem

6 (4+4+4=12 pts.) (a) Find a linear combination \(w\) of the linearly independent vectors \(v\) and \(u\) that is perpendicular to \(u\).

(b) For the 2-column matrix \(A = \begin{bmatrix} u & v \end{bmatrix}\), find \(Q\) (orthonormal columns) and \(R\) (2 by 2 upper triangular) so that \(A = QR\).

(c) In terms of \(Q\) only, using \(A = QR\), find the projection matrix \(P\) onto the plane spanned by \(u\) and \(v\).

Solution (a) You could just write down \(w = 0u + 0v = 0\)—that's perpendicular to everything! But a more useful choice is to subtract off just enough \(u\) so that \(w = v - cu\) is perpendicular to \(u\). That means \(0 = w^Tu = v^Tu - cu^Tu\), so \(c = (v^Tu)/(u^Tu)\) and \[ w = v - \left( \frac{v^Tu}{u^Tu} \right)u. \] (b) We already know \(u\) and \(w\) are orthogonal; just normalize them! Take \(q_1 = u/|u|\) and \(q_2 = w/|w|\). Then solve for the columns \(r_1, r_2\) of \(R\): \(Qr_1 = u\) so \(r_1 = \begin{bmatrix} |u| \\ 0 \end{bmatrix}\), and \(Qr_2 = v\) so \[ r_2 = \begin{bmatrix} c|u| \\ |w| \end{bmatrix} \quad \text{where } c = \frac{v^Tu}{u^Tu} \text{ as before.} \] Then \(Q = \begin{bmatrix} q_1 & q_2 \end{bmatrix}\) and \(R = \begin{bmatrix} r_1 & r_2 \end{bmatrix}\). (c) \(P = A(A^TA)^{-1}A^T = (QR)(R^TQ^TQR)^{-1}(R^TQ^T) = (QR)(R^TQ^T) = QQ^T\).

Keywords: Gram-Schmidt, QR factorization, orthogonal vectors, projection