{%hackmd 5xqeIJ7VRCGBfLtfMi0_IQ %} # Singular value decomposition ## Problem Let $\alpha = \{\bv_1, \bv_2, \bv_3\}$ with $$ \bv_1 = \frac{1}{\sqrt{3}}\begin{bmatrix} 1 \\ 1 \\ 1 \end{bmatrix},\ \bv_2 = \frac{1}{\sqrt{2}}\begin{bmatrix} 1 \\ -1 \\ 0 \end{bmatrix},\ \bv_3 = \frac{1}{\sqrt{6}}\begin{bmatrix} 1 \\ 1 \\ -2 \end{bmatrix}. $$ Let $\beta = \{\bu_1, \bu_2, \bu_3, \bu_4\}$ with $$ \bu_1 = \frac{1}{\sqrt{2}}\begin{bmatrix} 1 \\ 1 \\ 0 \\ 0 \end{bmatrix}, \ \bu_2 = \frac{1}{\sqrt{2}}\begin{bmatrix} -1 \\ 1 \\ 0 \\ 0 \end{bmatrix}, \ \bu_3 = \frac{1}{\sqrt{2}}\begin{bmatrix} 0 \\ 0 \\ 1 \\ 1 \end{bmatrix}, \ \bu_4 = \frac{1}{\sqrt{2}}\begin{bmatrix} 0 \\ 0 \\ -1 \\ 1 \end{bmatrix}. $$ Let $A$ be a $4\times 3$ matrix such that $$ [f_A]_\alpha^\beta = \begin{bmatrix} 3 & 0 & 0 \\ 0 & 2 & 0 \\ 0 & 0 & 0 \\ 0 & 0 & 0 \end{bmatrix}. $$ For $$ \bx = \begin{bmatrix} 0 \\ 1 \\ 2 \end{bmatrix}, $$ find $A\bx$. ## Thought Recall that $[f_A]_\alpha^\beta [\bx]_\alpha = [A\bx]_\beta$. Therefore, we need to transform $\bx$ into $[\bx]_\alpha$ and then transform $[A\bx]_\beta$ into $A\bx$ to get the answer. ## Sample answer Note that $\alpha$ is an orthonormal basis, so we have $$ \bx = c_1\bv_1 + c_2\bv_2 + c_3\bv_3 $$ with $c_i = \inp{\bx}{\bv_i}$ for $i = 1,2,3$. By direct computation, we have $$ \bx = \sqrt{3}\bv_1 - \frac{1}{\sqrt{2}}\bv_2 - \frac{3}{\sqrt{6}}\bv_3 \text{ and } [\bx]_\alpha = \begin{bmatrix} \sqrt{3} \\ -\frac{1}{\sqrt{2}} \\ -\frac{3}{\sqrt{6}} \end{bmatrix}. $$ Now we may compute $$ [A\bx]_\beta = [f_A]_\alpha^\beta [\bx]_\alpha = \begin{bmatrix} 3\sqrt{3} \\ -\sqrt{2} \\ 0 \\ 0 \end{bmatrix}. $$ This means $$ \begin{aligned} A\bx &= 3\sqrt{3}\bu_1 -\sqrt{2}\bu_2 + 0\bu_3 + 0\bu_4 \\ &= \frac{3\sqrt{3}}{\sqrt{2}}\begin{bmatrix} 1 \\ 1 \\ 0 \\ 0 \end{bmatrix} - \begin{bmatrix} -1 \\ 1 \\ 0 \\ 0 \end{bmatrix} = \begin{bmatrix} \frac{3\sqrt{3}}{\sqrt{2}} + 1 \\ \frac{3\sqrt{3}}{\sqrt{2}} - 1 \\ 0 \\ 0 \end{bmatrix}. \end{aligned} $$ *This note can be found at Course website > Learning resources.*