I don't quite understand how to get the convex conjugate of a function.

Wikipedia says the following

The convex conjugate of an affine function $$ f(x) = \left\langle a,x \right\rangle - b,\, a \in \mathbb{R}^n, b \in \mathbb{R} $$ is $$f^\star\left(x^{*} \right) = \begin{cases} b, & x^{*} = a \\ +\infty, & x^{*} \ne a. \end{cases} $$

But why does the $+\infty$ appear there?

So I know the definition of the convex conjugate $f^*$:

$$ f^*(x^*) = \sup_{x \in \mathbb{R}^n} \ \langle x^*, x \rangle -f(x) $$

then I found the derrivative ${f^*}'$

$$ {f^*}'(x^*) = x^* - a $$,

so my solution would be

$$ f^*(x^*) = b $$

Can someone shed light into this?


Applying the definition, if $f(x) = \langle a,x \rangle - b$, then $$f^*(x^*) = \sup_{x \in \mathbb R^n} \Big(\langle x^*, x\rangle - \langle a,x \rangle + b\Big) = b + \sup_{x \in \mathbb R^n} \langle x^*-a, x\rangle.$$ In the case when $x^* = a$, this is just $b + \sup_{x \in \mathbb{R}^n} 0 = b$. But if $x^* \ne a$, we can choose $x$ in the sup to be $C(x^* - a)$ for any $C$. Therefore $$f^*(x^*) \ge b + \langle x^* - a, C(x^* - a)\rangle = b + C\|x^*-a\|_2^2.$$ Since $\|x^*-a\|_2^2 > 0$, we can make this lower bound on $f^*$ as high as we like by choosing very very large $C$, so $f^*(x^*) = +\infty$.


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.