Skip to content

3.5-final-2005-strang

Final Exam, 2005 (strang) · Section 3.5: Dimensions of the Four Subspaces Source: finalexam_1.pdf, p. 124

Problem

3 (10 pts.) (a) I was looking for an \(m\) by \(n\) matrix \(A\) and vectors \(b, c\) such that \(Ax = b\) has no solution and \(A^Ty = c\) has exactly one solution. Why can I not find \(A, b, c\)?

(b) In \(R^m\), suppose I gave you a vector \(b\) and a vector \(p\) and \(n\) linearly independent vectors \(a_1, a_2, \dots, a_n\). If I claim that \(p\) is the projection of \(b\) onto the subspace spanned by the \(a\)'s, what tests would you make to see if this is true?

Solution (a) If \(Ax = b\) has no solution, the column space of \(A\) must have dimension less than \(m\). The rank is \(r < m\). Since \(A^Ty = c\) has exactly one solution, the columns of \(A^T\) are independent. This means that the rank of \(A^T\) is \(r = m\). This contradiction proves that we cannot find \(A\), \(b\) and \(c\). (b) We need to check two statements: the vector \(b - p\) is orthogonal to the space generated by \(a_1, \dots, a_n\), and the vector \(p\) lies in that subspace. The first condition we check by seeing if the scalar products \(a_1 \cdot (b - p), \dots, a_n \cdot (b - p)\) all equal zero. The second condition we check by considering the \((n+1) \times m\) matrix whose first \(n\) rows are the coordinates of the \(a_i\)'s and whose last row consists of the coordinates of \(p\). The vector \(p\) is in the span of the \(a_i\)'s if and only if the last row becomes zero in elimination.

Keywords: rank, four subspaces, projection, orthogonality