4.2-exam2-2012-strang¶
Quiz 2, 2012 (strang) · Section 4.2: Projections onto Lines and Subspaces Source: exam2_2.pdf, pp. 29-30
Problem¶
1 (40 pts.)
(a) Find the projection \(p\) of the vector \(b\) onto the plane of \(a_1\) and \(a_2\), when
\[
b = \begin{bmatrix} 1 \\ 0 \\ 0 \\ 1 \end{bmatrix}, \quad a_1 = \begin{bmatrix} 1 \\ 7 \\ 1 \\ 7 \end{bmatrix}, \quad a_2 = \begin{bmatrix} -1 \\ 7 \\ 1 \\ -7 \end{bmatrix}.
\]
(b) What projection matrix \(P\) will produce the projection \(p = Pb\) for every vector \(b\) in \(R^4\)?
(c) What is the determinant of \(I - P\)? Explain your answer.
(d) What are all nonzero eigenvectors of \(P\) with eigenvalue \(\lambda = 1\)? How is the number of independent eigenvectors with \(\lambda = 0\) of a square matrix \(A\) connected to the rank of \(A\)?
(You could answer (c) and (d) even if you don't answer (b).)
Solution
(a) Observe that \(a_1^Ta_2 = 0\). Thus \[ p = \frac{a_1^Tb}{a_1^Ta_1}a_1 + \frac{a_2^Tb}{a_2^Ta_2}a_2 = \frac{8}{100}a_1 - \frac{8}{100}a_2 = \begin{bmatrix} 4/25 \\ 0 \\ 0 \\ 28/25 \end{bmatrix}. \] (b) Let \(A\) be the \(4 \times 2\) matrix with columns \(a_1, a_2\). \(P\) is given by \(P = A(A^TA)^{-1}A^T\). Notice that \[ A^TA = \begin{bmatrix} 100 & 0 \\ 0 & 100 \end{bmatrix} \] (\(a_1\) and \(a_2\) are orthogonal and of the same length). Thus \[ P = \frac{1}{100}AA^T = \frac{1}{100} \begin{bmatrix} 2 & 0 & 0 & 14 \\ 0 & 98 & 14 & 0 \\ 0 & 14 & 2 & 0 \\ 14 & 0 & 0 & 98 \end{bmatrix}. \] # corrected source typo: printed solution says dim N(A^T) = 3; A is 4×2 of rank 2, so dim N(A^T) = 4−2 = 2, and rank(I−P) = 2 (c) \(I - P\) is the matrix of the projection onto the orthogonal complement of \(C(A)\), i.e., \(N(A^T)\). In particular, \(I - P\) has rank equal to the dimension of \(N(A^T)\), which is 2. Thus \(I - P\) is singular, and \(\det(I - P) = 0\). (d) The nonzero eigenvectors with eigenvalue \(\lambda = 1\) are all the nonzero linear combinations of \(a_1\) and \(a_2\), i.e., all the nonzero vectors in the plane spanned by \(a_1\) and \(a_2\). For an \(n \times n\) matrix \(A\) of rank \(r\), the number of independent eigenvectors with eigenvalue 0 equals the number of independent vectors in \(N(A)\), which is the dimension of \(N(A)\), namely \(n - r\).Keywords: projection matrix, orthogonal vectors, determinant, eigenvectors, rank