Prove that triangle's inequality in the definition of a norm is equivalent to the convexity of the closed unit ball.

(I proved triangle inequality implies convexity, but I could not prove converse. For converse I tried like this: I took any $x$, $y$ in the normed space $X$, then I derived two new unit vectors $x/||x||$, $y/||y||$ so that the new vectors lie in the closed unit ball and so I can use the convexity of the ball for the vectors $x$ & $y$. But further I could not get the result this way.)

  • $\begingroup$ Do you have any attempts to share? Perhaps what you have tried, you could post what you know and people will be able to help you from there. $\endgroup$ – smokeypeat Aug 8 '17 at 10:55

Triangle's inequality in the definition of a norm is as follows:

$$\forall x,y\in X,\ \|x+y\|\le \|x\|+\|y\| $$

Now, the above inequality follows the convexity of the unit closed ball. Let $\lambda\in [0,1]$ and $\mathbf{x},\mathbf{y}\in \mathbb{B}[0,1]\implies \|\mathbf{x}\|\le 1 $ and $\|\mathbf{y}\|\le 1$. Now consider, \begin{align*} \|\lambda \mathbf{x}+(1-\lambda) \mathbf{y}\| & \le \|\lambda \mathbf{x}\|+\|(1-\lambda)\mathbf{y}\|\\ & = \lambda \|\mathbf{x}\|+(1-\lambda)\|\mathbf{y}\|\\ & \le \lambda + (1-\lambda)=1 \end{align*} Hence, $ \lambda \mathbf{x}+(1-\lambda) \mathbf{y}\in \mathbb{B}[0,1]. $

For the converse see, $$ \frac{\mathbf{x}+\mathbf{y}}{\|\mathbf{x}\|+\|\mathbf{y}\|}=\frac{\|\mathbf{x}\|}{\|\mathbf{x}\|+\|\mathbf{y}\|}\frac{\mathbf{x}}{\|\mathbf{x}\|}+\frac{\|\mathbf{y}\|}{\|\mathbf{x}\|+\|\mathbf{y}\|}\frac{\mathbf{y}}{\|\mathbf{y}\|}\in \mathbb{B}[0,1] \implies \|\mathbf{x}+\mathbf{y}\|\le \|\mathbf{x}\|+\|\mathbf{y}\|$$


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.