4.3-exam2-1999-fall¶
Quiz 2, 1999 (fall) · Section 4.3: Least Squares Approximations Source: exam2_1.pdf, pp. 31, 35
Problem¶
1 (25 pts.) (a) Find equations (do not solve) for the coefficients \(C, D, E\) in \(b = C + Dt + Et^2\), the parabola which best fits the four points
\[
(t, b) = (0, 0), (1, 1), (1, 3) \text{ and } (2, 2).
\]
(b) In solving this problem you are projecting the vector \(b =\) __ onto the subspace spanned by __. The projection in terms of \(C, D, E\) is \(p =\) ___.
Solution
1a. Let \[ A = \begin{bmatrix} 1 & 0 & 0 \\ 1 & 1 & 1 \\ 1 & 1 & 1 \\ 1 & 2 & 4 \end{bmatrix}, \quad b = \begin{bmatrix} 0 \\ 1 \\ 3 \\ 2 \end{bmatrix}. \] \(Ax = b\) has no solution, so we look for its least squares solution and solve the system \[ A^T A \begin{bmatrix} C \\ D \\ E \end{bmatrix} = A^T b, \quad \text{i.e., } \begin{bmatrix} 4 & 4 & 6 \\ 4 & 6 & 10 \\ 6 & 10 & 18 \end{bmatrix} \begin{bmatrix} C \\ D \\ E \end{bmatrix} = \begin{bmatrix} 6 \\ 8 \\ 12 \end{bmatrix}. \] 1b. You are projecting the vector \(b = (0, 1, 3, 2)\) onto the subspace spanned by the column vectors of \(A\). The projection in terms of \(C, D, E\) is \[ p = A\hat{x} = A \begin{bmatrix} C \\ D \\ E \end{bmatrix} = \begin{bmatrix} C \\ C + D + E \\ C + D + E \\ C + 2D + 4E \end{bmatrix}. \]Keywords: least squares, parabola fit, normal equations, projection