Strengthened converses of the Jensen and Edmundson-Lah-Ribarič inequalities (Q2520623)

Assume that \(E\) is a nonempty set and let \(L\) be a vector space of real-valued functions \(f:E\to\mathbb{R}\) having following properties: {\parindent=0.7cm \begin{itemize}\item[(1)] \(f, g \in L \Rightarrow af + bg \in L\) for all \(a, b \in\mathbb{R}\); \item[(2)] \(1 \in L\), where \(1\) denotes the constant function; \item[(3)] \(f, g \in L\Rightarrow \min\{f, g\} \in L \wedge \max\{f, g\}\in L\). \end{itemize}} In other words, \(L\) is a subspace of the vector space \(\mathbb{R}^E\) over \(\mathbb{R}\) containing \(1\) such that the maximum and the minimum of two elements of \(L\) is in it. For positive linear functionals \(A : L \to \mathbb{R}\), the authors say that \(A\) is a positive normalized linear functional if \(A(1) = 1\). Let \(A\) be a positive normalized linear functional and let \(\varphi:I\to\mathbb{R}\) be a continuous convex function. Then the generalization of the Jensen inequality says that, for all \(f\in L\) such that \(\varphi(f)\in L\), it follows that \(A(f)\in I\) and \[ \varphi(A(f))\leq A(\varphi(f)). \] In the paper under review, the authors state some very nice variant converses of Jensen's inequality. Also, in continuation, the authors give many applications; for instance, to generalized means, the Hölder inequality, the Hermite-Hadamard's inequality, and the Giaccardi and Petrović inequalities.