Convex conjugate
In mathematics and mathematical optimization, the convex conjugate of a function is a generalization of the Legendre transformation which applies to non-convex functions. It is also known as Legendre–Fenchel transformation, Fenchel transformation, or Fenchel conjugate. The convex conjugate is widely used for constructing the dual problem in optimization theory, thus generalizing Lagrangian duality.
Definition
Let be a real topological vector space and let be the dual space to. Denote bythe canonical dual pairing, which is defined by
For a function taking values on the extended real number line, its is the function
whose value at is defined to be the supremum:
or, equivalently, in terms of the infimum:
This definition can be interpreted as an encoding of the convex hull of the function's epigraph in terms of its supporting hyperplanes.
Examples
For more examples, see.- The convex conjugate of an affine function is
- The convex conjugate of a power function is
- The convex conjugate of the absolute value function is
- The convex conjugate of the exponential function is
Connection with expected shortfall (average value at risk)
SeeLet F denote a cumulative distribution function of a random variable X. Then,
has the convex conjugate
Ordering
A particular interpretation has the transformas this is a nondecreasing rearrangement of the initial function f; in particular, for f nondecreasing.
Properties
The convex conjugate of a closed convex function is again a closed convex function. The convex conjugate of a polyhedral convex function is again a polyhedral convex function.Order reversing
Declare that if and only if for all Then convex-conjugation is order-reversing, which by definition means that if thenFor a family of functions it follows from the fact that supremums may be interchanged that
and from the max–min inequality that
Biconjugate
The convex conjugate of a function is always lower semi-continuous. The biconjugate is also the closed convex hull, i.e. the largest lower semi-continuous convex function withFor proper functions
Fenchel's inequality
For any function and its convex conjugate, Fenchel's inequality holds for every andFurthermore, the equality holds only when, where is the subgradient.
The proof follows from the definition of convex conjugate:
Convexity
For two functions and and a number the convexity relationholds. The operation is a convex mapping itself.
Infimal convolution
The infimal convolution of two functions and is defined asThe operation is symmetric and associative, i.e.
Let be proper, convex and lower semicontinuous functions on Then the infimal convolution is convex and lower semicontinuous, and satisfies
or, equivalently,
which expresses the behaviour of convex conjugation with respect to sums of functions.
The infimal convolution of two functions has a geometric interpretation: The epigraph of the infimal convolution of two functions is the Minkowski sum of the epigraphs of those functions.
Maximizing argument
If the function is differentiable, then its derivative is the maximizing argument in the computation of the convex conjugate:hence
and moreover
Scaling properties
If for some , thenBehavior under linear transformations
Let be a bounded linear operator. For any convex function onwhere
is the preimage of with respect to and is the adjoint operator of
A closed convex function is symmetric with respect to a given set of orthogonal linear transformations,
if and only if its convex conjugate is symmetric with respect to