Efficiency and the uniform linear minorization of convex functions (Q688957): Difference between revisions

In a Banach space \(X\) with the unit ball \(B\) a family \(\{f_ i: i\in I\}\) of proper lower semicontinuous convex functionals is considered. The often used assumption concerning the variational convergence of such a family is the so-called uniform minorization of this family e.g. like \[ \exists\mu\geq 0\quad \forall i\in I\quad \forall x\in X:\quad f_ i(x)\geq-\mu(1+| x|).\tag{*} \] The author gives necessary and sufficient conditions for the uniform minorization of the form (*). The following theorem characterizes the main result. Theorem 3.4: Let \(\{f_ i: i\in I\}\) be a family of proper lower semicontinuous convex functionals on the Banach space \(X\). Then the following statements are equivalent: i) (*) holds and \(\{\text{epi}; f_ i: i\in I\}\) is efficient (the last means: there is a bounded set \(A\subset X\times I\negthinspace R\) such that \(A\cap\text{epi}; f_ i\neq\emptyset\) for all \(i\in I\)). ii) There are \(\delta>0\), \(\rho>0\) and for each \(i\in I\) a point \(x_ i\in\rho B\) such that \(f_ i(x_ i)\leq\rho\) and \(\inf_{x\in X_ i+ \delta U} f_ i(x)\geq-\rho\). iii) The family \(\{\Delta(f_ i): i\in I\}\) is efficient, where \[ \begin{aligned} \Delta(f_ i) & :=\{(x,f_ i(x), y^*)\in X\times I\negthinspace R\times X^*;\;f_ i^*(y^*)+ f_ i(x)=\langle y^*,x\rangle\} \text{ and}\\ f_ i^*(y^*) & := \sup_{x\in X}(\langle y^*,x\rangle-f_ i(x)).\end{aligned} \] The main tool is the so-called Borwein-variational principle being used during the proof of ii) \(\to\) iii). Simple examples show that the assumptions of the theorem are essential. Applying Theorem 3.4 ii) it is proved that the Mosco convergence implies the property (*).