I need to prove the following:

Let $A \in \mathbf R^{n*n}$ be a real matrix and $x \in \mathbf R^n$ a vector.show that: $$\Vert Ax \Vert_{2} \geq s_{\min}\Vert x \Vert_{2}$$ where $s_{\min}$ is the smallest singular value of A.

I proceeded like so:

per def: $\Vert A \Vert_{2} = \max\left\{ \sqrt w;\text{w is the EV of $A^TA$} \right\}$

so we see that $$\Vert A \Vert_{2} \geq s_{\min}$$ multiply each side by $\Vert x\Vert_{2}$ $$\Vert A\Vert_{2} \Vert x \Vert_{2} \geq s_{ \min}\Vert x \Vert_{2}$$ can we now conclude from this that $$\Vert Ax \Vert_{2} \geq s_{\min}\Vert x \Vert_{2}$$??

since $$\Vert A\Vert_{2} \Vert x \Vert_{2} \geq \Vert Ax \Vert_{2}$$ this doesn't seem quite right to me! please help.

Thanks a lot in advance


Note that $\|Ux\| = \|x\|$ for an orthogonal $U$.

Hence $\|Ax\| = \|U \Sigma V^T x \| = \|\Sigma V^T x \| = \sqrt{ \sum_l \sigma_k^2 [V^T x]_k^2 } \ge \sigma_n \|V^T x\| = \sigma_n \|x\|$.



When $A$ is singular its smallest singular value is zero and thus the inequality is trivial this case.


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.