4.1-exam2-2007-johnson¶
Quiz 1, 2007 (johnson) · Section 4.1: Orthogonality of Vectors and Subspaces Source: exam2_1.pdf, p. 151
Problem¶
2 (30 pts.) Fill in the blanks below: (You don't need to justify your answer.)
(a) The nullspace of \(AB\) contains the nullspace of ______.
(b) Let \(P\) be the projection matrix to the row space of a matrix \(A\), then \(I - P\) is the projection matrix to ______.
(c) Suppose \(A\) is an \(m \times n\) matrix, and the row space of \(A\) is \(n\) dimensional, then its nullspace is ______ dimensional.
(d) Let \(\hat{x}\) be the least-squares solution to \(Ax = b\). Then \(b - A\hat{x}\) is orthogonal to the ______ of \(A\).
(e) If \(A^T = -A\) (\(A\) is antisymmetric), and \(A\) is \(n \times n\) where \(n\) is odd, then \(\det A = \_\_\_\_\_\_\).
(f) If \(A\) is symmetric and \(P\) is the projection matrix onto the nullspace \(N(A)\), then \(PA = \_\_\_\_\_\_\).
Solution
(a) \(B\). If \(Bx = 0\) then \(ABx = 0\). (b) \(N(A)\). Reason: \(I - P\) is the projection onto the orthogonal complement, and the orthogonal complement of the row space is the nullspace. (c) 0. The rank \(r\) of \(A\) is the dimension of the row space, so \(r = n\), and the nullspace has dimension \(n - r = n - n = 0\). (d) Column space. The least-squares solution solves \(A\hat{x} = Pb\), where \(P\) is the projection onto \(C(A)\), so \(b - A\hat{x} = b - Pb = (I - P)b\), and \(I - P\) projects onto the complement of \(C(A)\). Equivalently, the least-squares solution finds the closest point \(A\hat{x}\) to \(b\) in \(C(A)\), so the difference must be perpendicular to \(C(A)\). (e) 0. Reason: \(\det A = \det A^T = \det(-A) = (-1)^n \det A = -\det A\) since \(n\) is odd, and the only way to have \(\det A = -\det A\) is if \(\det A = 0\). (f) 0. Since \(A\) is symmetric, \(N(A) = N(A^T) = C(A)^{\perp}\), so \(P\) projects onto the orthogonal complement of \(C(A)\). Thus \(PA = 0\), since \(P\) projects every column of \(A\) to zero.Keywords: four subspaces, nullspace, projection matrix, determinant