Skip to content

4.3-exam1-2001-fall

Quiz 1, 2001 (fall) · Section 4.3: Least Squares Approximations Source: exam1_1.pdf, pp. 48-49

Problem

Problem 1. We want to solve the equation \(A\vec{x} = \vec{b}\) "as well as possible" when no exact solution exists.

(a) What does "as well as possible" mean?

(b) The first (and \(n\)-th) entries of \(A^T\vec{e}\), where \(\vec{e} = \vec{b} - A\vec{x}\), are zero. What does this mean?

Solution (a) "As well as possible" means that we find a (the) vector \(\vec{x}\) such that the vector \[ \vec{e} = \vec{b} - A\vec{x} \] has the least possible length. (b) The fact that the first (\(n\)-th) entry in \(A^T e\) is zero means that \(e\) is orthogonal to the first (\(n\)-th) row of \(A^T\) or, in other words, \(e\) is orthogonal to the each column of \(A\). Another way to put it is that \(e\) is orthogonal to the column space of \(A\).

Keywords: least squares, normal equations, error vector, orthogonal