1
$\begingroup$

The canonical definition of Fenchel conjugate function is that for a function $f(x)$, $$f^*(x^*) = \textsf{sup}_{x}\{\langle x^*, x \rangle - f(x)\}$$ which is equivalent to $f^*(x^*) =-\textsf{inf}_x\{f(x)-\langle x^*, x \rangle\}$. Recently as I was reading Optimization by vector space methods, I noticed that in Section 8.6, in order to prove the Lagrange duality theorem, the book introduces another dual functional: $$\textsf{inf}_{x}\{f(x)+ \langle x^*, x\rangle\}$$ The book claims that those two definitions are essentially the same. I can see that they have the same geometric interpretation, but I failed to prove that they are the same rigorously.

$\endgroup$
3
$\begingroup$

If you define $$f^*(y) = \textsf{sup}_{x}\{\langle x, y\rangle - f(x)\}$$ $$g(y) = \textsf{inf}_{x}\{f(x)+ \langle x, y\rangle\},$$ then $f^*(y) = -g(-y)$.

$\endgroup$

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.