Fenchel-Rockafellar duality: From Banach space to Hilbert space
Fenchel-Rockafellar duality theorem
Fenchel-Rockafellar duality: From Banach space to Hilbert space
Let B, H be respectively a Banach, Hilbert space. Let f:H⟶R, g:B⟶R and A:B⟶H. Consider the following ubiquitous primal problem x∈Bminf(Ax)+g(x)
It is well-known that the Fenchel-Rockafellar dual problem is u∈Hmax−f∗(−u)−g∗(A∗u)
where f∗:H⟶R, g∗:B∗⟶R and A∗:H⟶B∗. Here f∗,g∗ are convex conjugate of f and g, respectively, and A∗ is the adjoint operator of A.
In this post, we would like to show this result by using the change of variable.
Let v=Ax, the primal problem can be rewritten, assuming that A is surjective, i.e. ImA=H, v∈Hminf(v)+g^(v)
where g^(v):=inf{g(x):Ax=v}
By applying Fenchel duality problem, we obtain the following dual problem x∈Bmax−f∗(−u)−g^∗(u)
The rest is to show that g^∗(u)=g∗(A∗u)
Indeed, g^∗(u)=vsup⟨u,v⟩−g^(u)=xsup⟨u,Ax⟩−g(x)=g∗(A∗u)
We obtained desired result.
With this observation, we see that the second primal problem only involves the Hilbert space, which is more easier for theoretical analysis, for example, the optimality conditions.