Bounds for the normalised Jensen functional. (English) Zbl 1113.26021

This is a paper on bounds for the normalized Jensen functional. Jensen’s inequality for convex functions is one of the best known and extensively used inequalities in various fields. It is a source of many classical inequalities including the generalized triangle inequality, the arithmetic mean-geometric mean-harmonic mean inequality, the positivity of relative entropy, Shannon’s inequality, Levinson’s inequality and other results. Let
\[ \mathcal P_n : = \left\{ (p_1, p_2, \dots, p_n) \mid p_j \geq 0 \;(j=1, \dots, n),\;\sum_{j=1}^n p_j = 1\right\}. \]
The author considers the normalized Jensen functional
\[ \mathcal J_n(f, X, P): = \sum_{j=1}^n p_j f(x_j) - f\left(\sum_{j=1}^n p_j x_j\right) \geq 0 , \]
where \(f: C \rightarrow \mathbb R\) is a convex function on the convex set \(C\), \(X= (x_1, \dots, x_n) \in C^n\) and \(P=(p_1, \dots, p_n) \in \mathcal P_n\), and proves that if \(P, Q \in \mathcal P_n\), \(q_j >0 \;(j=1, \dots, n)\), then
\[ \max_{1\leq j\leq n} \{{p_j}/{q_j}\} \mathcal J_n(f, X, Q)\geq \mathcal J_n(f, X, P) \geq \min_{1\leq j\leq n}\{{p_j}/{q_j}\} \mathcal J_n(f, X, Q) \geq 0 . \]
As applications, a generalized arithmetic mean-geometric mean-harmonic mean inequality, the positivity of Kullback-Leibler divergence and a refinement of Shannon’s inequality are obtained.


26D15 Inequalities for sums, series and integrals
26D10 Inequalities involving derivatives and differential and integral operators
Full Text: DOI


[1] Pečarić, Convex functions, partial orderings and statistical applications (1992) · Zbl 0749.26004
[2] Mitrinović, Classical and new inequalities in analysis (1993)
[3] Bullen, Handbook of mean and their inequalities (2003) · Zbl 1035.26024
[4] Dragomir, Selected topics on Hermite-Hadamard inequalities and applications (2000)
[5] DOI: 10.1002/0471200611
[6] McEliece, The theory of information and coding (1977)
This reference list is based on information provided by the publisher or from digital mathematics libraries. Its items are heuristically matched to zbMATH identifiers and may contain data conversion errors. It attempts to reflect the references listed in the original paper as accurately as possible without claiming the completeness or perfect precision of the matching.