4
$\begingroup$

It is well known in convex analysis that when a closed, proper, function $f$ is Legendre-type, that is, essentially strictly convex and essentially smooth, the Legendre transform yields a dual function $g$ which has the same properties, whereby the gradient is a bijection from the domain of $f$ to the domain of $g$.

Does there exist a similar set of results for strongly convex $f$? or for essentially strongly convex $f$?

$\endgroup$

1 Answer 1

3
$\begingroup$

Yes, strong convexity is conjugate to uniform smoothness or Lipschitz-continuous differentiability (where the Lipschitz constant is the reciprocal of the modulus of strong continuity), see, e.g.,

Azé, Dominique; Penot, Jean-Paul, Uniformly convex and uniformly smooth convex functions, Ann. Fac. Sci. Toulouse, VI. Sér., Math. 4, No. 4, 705-730 (1995). ZBL0870.49010.

$\endgroup$
2
  • $\begingroup$ Does the gradient remain a bijection? If not, what guarantees it? $\endgroup$
    – Concu Bine
    Sep 6, 2019 at 1:37
  • 1
    $\begingroup$ Yes, strongly convex functions are in particular strictly convex, which is the property that guarantees the bijection (of the subgradient as a set-valued operator, which reduces to the gradient if you also have smoothness). $\endgroup$ Sep 6, 2019 at 6:34

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.