Skip to content

4.2-exam2-2009

Quiz 2, 2009 · Section 4.2: Projections onto Lines and Subspaces Source: exam2_1.pdf, pp. 165-166

Problem

Problem 1 (20 points = 5 + 5 + 5 + 5): fill in the blanks.

(a) If \(P\) is the projection matrix onto the null space of \(A\), then \(Py - y\), for any \(y\), is in the ___ space of \(A\).

(b) If \(Ax = b\) has a solution \(x\), then the closest vector to \(b\) in \(N(A^T)\) is ___ (best answer).

(c) If the rows of \(A\) (an \(m \times n\) matrix) are independent, then the dimension of \(N(A^T A)\) is ___.

(d) If a matrix \(U\) has orthonormal rows, then \(I = \_\_\_\) and the projection matrix onto the row space of \(U\) is ___. (Your answers should be the simplest expressions involving \(U\) and \(U^T\) only.)

Solution Answers: (a) row; (b) 0; (c) \(n - m\); (d) \(UU^T\), \(U^T U\). (a) Since \(Py\) is the projection of \(y\) onto the null space of \(A\), \(Py - y\) is orthogonal to the null space; it then must lie in the row space of \(A\). (b) Since \(Ax = b\) has a solution \(x\), \(b\) is in the column space \(C(A)\) of \(A\). The left null space \(N(A^T)\) is orthogonal to the column space, so the closest vector to \(b\) in \(N(A^T)\) is 0. (c) The null space of \(A^T A\) is the same as the null space of \(A\); the latter has dimension \(n - m\) because \(A\) is of full row rank \(m\). (Equivalently: the ranks of \(A\) and \(A^T A\) are the same, both equal to \(m\), and since \(A^T A\) has \(n\) columns, \(N(A^T A)\) has dimension \(n - m\).) (d) Note that \(U^T = Q\), a matrix with orthonormal columns. As seen in class, \(I = Q^T Q = UU^T\), and the projection matrix onto \(C(Q) = C(U^T)\) is \(QQ^T = U^T U\).

Keywords: projection matrix, nullspace, row space, closest vector