Most convex functions have unique minimizers (Q2821976)

The space of classes of proper lower semi-continuous convex functions \(f:\mathbb R^n\to\mathbb R\) is considered. Functions with constant difference are identified into equivalence classes \([f]\). Two representations of this space are considered. Each class can be identified with the subdifferential \(\partial f\) of any of its elements, and each subdifferential can be identified with the proximal mapping which is defined in terms of the Morreau envelope and fulfills the identity \(P_1 f=(\partial f+\mathrm{Id})^{-1}\). The convex classes of functions with the unique minimizer correspond to the subdifferentials having unique zero point, and to the proximal functions with unique fixed point. A complete metric is introduced which agrees with the locally uniform convergence of \(P_1f\)'s.NEWLINENEWLINEThe main result is that the set of proximal mappings with unique fixed point is generic (residual) with respect to the above mentioned complete metric. The same result is deduced in terms of subdifferentials and their distances in the Attouch-Wets distance which agrees with the graph convergence.