Skip to content

4.4-final-2014-strang

Final Exam, 2014 · Section 4.4: Orthonormal Bases and Gram-Schmidt Source: finalexam_2.pdf, p. 136

Problem

3 (10 points)

(a) Suppose \(q_1, q_2\) are orthonormal in \(\mathbb{R}^4\), and \(v\) is NOT a combination of \(q_1\) and \(q_2\). Find a vector \(q_3\) by Gram-Schmidt, so that \(q_1, q_2, q_3\) is an orthonormal basis for the space spanned by \(q_1, q_2, v\).

(b) If \(p\) is the projection of \(b\) onto the subspace spanned by \(q_1\) and \(q_2\) and \(v\), find \(p\) as a combination of \(q_1, q_2, q_3\). (You are solving the least squares problem \(Ax = b\) with \(A = [q_1, q_2, q_3]\).)

Solution (a) Subtract the components of \(v\) along the orthonormal vectors \(q_1\) and \(q_2\), then normalize: \[ q_3 = \frac{v - (v^Tq_1)q_1 - (v^Tq_2)q_2}{\|v - (v^Tq_1)q_1 - (v^Tq_2)q_2\|}. \] The numerator is nonzero precisely because \(v\) is not a combination of \(q_1\) and \(q_2\), and \(q_3\) is orthogonal to \(q_1\) and \(q_2\) by construction. (b) Since the columns of \(A = [q_1, q_2, q_3]\) are orthonormal, \(A^TA = I\) and the projection formula simplifies: \[ p = A(A^TA)^{-1}A^Tb = AA^Tb = q_1(q_1^Tb) + q_2(q_2^Tb) + q_3(q_3^Tb). \] The projection of \(b\) onto the subspace is just the sum of its components along the orthonormal basis vectors.

Keywords: Gram-Schmidt, orthonormal basis, projection, least squares