2
$\begingroup$

Find a singular value decomposition for

$A$ = \begin{bmatrix}1&1&0\\1&0&1\end{bmatrix}

I know that the steps of finding an SVD for a matrix $A$ such that $A$ = $U$$\sum$$V^T$ are the following:

1) Find $A^T$$A$.

2) Find the eigenvalues of $A^T$$A$.

3) Find the eigenvectors of $A^T$$A$.

4) Set up $\sum$ using the positive eigengalues of $A^T$$A$, placing them in a diagonal matrix using the format of the original matrix $A$, with $0$ in all the other entries.

5) Normalize the eigenvectors of $A^T$$A$ to get the matrix $V$ (which will need to be transposed).

6) Find $U$ using the normalized eigenvectors of $V$, where ui = $\frac{1}{σ_{i}}$$A$vi, and where vi are the columns of $V$.

However, after (1), I get \begin{bmatrix}2&1&1\\1&1&0\\1&0&1\end{bmatrix}. Then, after using Row $3$ to find the eigenvalues, I get $(-1-λ) + (1-λ) [(2-λ)(1-λ)-1]$. This simplifies to $(-1+λ)-(λ^3-4λ^2+4λ+1)$.

Is this the correct way to go about this? This is Elementary Linear Algebra, yet just finding the eigenvalues and eigenvectors is quite drawn-out.

Any suggestions or tips would be helpful.

$\endgroup$
0
$\begingroup$

I think there is a mistake in your calculation. Notice that:

$$ \det(A^\top A - \lambda I) = \det\begin{pmatrix} 2 - \lambda & 1 & \fbox{$1$} \\ 1 & 1 - \lambda & 0 \\ 1 & 0 & \fbox{$1 - \lambda$} \end{pmatrix} = (1 - \lambda)\left[ (2 - \lambda)(1 - \lambda) - 1) \right] + 1 \cdot (0 -(1 - \lambda)) = (1 - \lambda)(2 + \lambda^2 - 3 \lambda - 2) = (1 - \lambda)(\lambda^2 - 3 \lambda) = \lambda (1 - \lambda)(\lambda - 3) $$ Therefore, you get $\lambda_1 = 0, \lambda_2 = 1, \lambda_3 = 3$.

$\endgroup$
0
$\begingroup$

Guide:

Rather than a $3$ by $3$ matrix, deal with a $2$ by $2$ matrix, $$AA^T= \begin{bmatrix} 1 & 1 & 0 \\ 1 & 0 & 1\end{bmatrix}\begin{bmatrix} 1 & 1 \\ 1 & 0\\ 0 & 1\end{bmatrix}= \begin{bmatrix} 2 & 1 \\ 1 & 2 \end{bmatrix}=U\Sigma U^T$$

$$(2-\lambda)^2-1=0$$ $$(1-\lambda)(3-\lambda)=0$$

Eigenvector corresponding to eigenvalue $1$ is $\frac1{\sqrt2}\begin{bmatrix} 1 \\ -1\end{bmatrix}$ and eigenvector corresponding to eigenvalue $3$ is $\frac1{\sqrt2}\begin{bmatrix} 1 \\ 1 \end{bmatrix}.$

$$A=U\Sigma V^T$$

Multiply $u_i^T$ on both sides, $u_i^TA=\sigma_i v_i^T$ and you can get $v_1^T$ and $v_2^T$. Remember that $V^T$ is an orthogonal matrix and we can use that property to obtain $v_3^T$.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.