Skip to content

3.4-final-2006

Final Exam, 2006 · Section 3.4: Independence, Basis, and Dimension Source: finalexam_1.pdf, p. 144

Problem

4 (3+4+5=12 pts.) Suppose \(A\) is a 5 by 3 matrix and \(Ax\) is never zero (except when \(x\) is the zero vector).

(a) What can you say about the columns of \(A\)?

(b) Show that \(A^TAx\) is also never zero (except when \(x = 0\)) by explaining this key step:

If \(A^TAx = 0\) then obviously \(x^TA^TAx = 0\) and then (WHY?) \(Ax = 0\).

(c) We now know that \(A^TA\) is invertible. Explain why \(B = (A^TA)^{-1}A^T\) is a one-sided inverse of \(A\) (which side of \(A\)?). \(B\) is NOT a 2-sided inverse of \(A\) (explain why not).

Solution (a) \(N(A) = 0\) so \(A\) has full column rank \(r = n = 3\): the columns are linearly independent. (b) \(x^TA^TAx = (Ax)^TAx\) is the squared length of \(Ax\). The only way it can be zero is if \(Ax\) has zero length (meaning \(Ax = 0\)). (c) \(B\) is a left inverse of \(A\), since \(BA = (A^TA)^{-1}A^TA = I\) is the (3-by-3) identity matrix. \(B\) is not a right inverse of \(A\), because \(AB\) is a 5-by-5 matrix but can only have rank 3. (In fact, \(AB = A(A^TA)^{-1}A^T\) is the projection onto the (3-dimensional) column space of \(A\). The printed solutions mistakenly write \(BA\) here.)

Keywords: independent columns, normal equations, left inverse, projection