3

I have to show that $2\Vert{A}\Vert$ is the smallest Lipschitz constant of $\nabla f$, where $f(x)=x^{T}Ax+2b^{T}x+c$.

$\forall x,y\in \mathbb{R}^n, \Vert\nabla f(x)-\nabla f(y)\Vert = \Vert2(Ax+b)-2(Ay+b)\Vert = \Vert2A(x-y)\Vert \le \Vert2A\Vert \Vert x-y\Vert$

but I don't know how to show that it's the smallest.

Mina
  • 576

2 Answers2

2

With the inequality you mentioned in the question, you proved that $2\Vert A\Vert$ is larger than the smallest Lipschitz constant.

Now I imagine that the matrix norm you’re using is the sup norm. That is

$$\Vert A \Vert =\sup\limits_{\Vert x \Vert =1} \Vert Ax\Vert$$ You can easily conclude using this definition as for any $c \lt \Vert A\Vert $ it exists $z$ with $\Vert z \Vert =1$ and $$\Vert Az\Vert \ge c.$$ Then use the linearity of $A$.

2

You can take $y = 0$ and then maximize over all $x$ with $||x|| = 1$ to get that $2||A||$ is the best constant.

Zarrax
  • 44,950