The stable Farkas lemma for composite convex functions in infinite dimensional spaces (Q2516052): Difference between revisions

The stable Farkas lemma, a generalization of the Farkas lemma, states that for every \(x^*\in \mathbb{R}^n\) and for each \(\alpha\in \mathbb{R},\) \[ [-g(x)\in S\Longrightarrow f(x)\geq \langle x^*,x\rangle +\alpha] \Longleftrightarrow (\exists \lambda\in S^+)(\forall x\in \mathbb{R}^n)\,f(x)+(\lambda\circ g)(x)\geq \langle x^*,x\rangle +\alpha, \] where \(S\subset \mathbb{R}^m\) is a closed convex cone with the dual cone \(S^+\), \(f:\mathbb{R}^n\to \mathbb{R}\cup \{+\infty\}\) is a given real-valued convex function and \(g:\mathbb{R}^n\to \mathbb{R}^m\) is a continuous convex function. In this paper a general composite convex optimization problem with a cone-convex system in a locally convex Hausdorff topological space is considered. The authors derive the Fenchel transform of some composite convex functions in order to prove a characterization of the stable Farkas lemma, expressed in terms of the epigraphs of the conjugates of the convex functions involved. The equivalent condition turns out to be weaker than the classical Slater condition. Some necessary and sufficient conditions for stable duality results of the composite convex functions are therefore obtained. In particular, it is shown that the \(K\)-monotonicity condition of the outer convex function \(f\) cannot be omitted.