Stolarsky means and Hadamard's inequality (Q1269634)

Let \(x,y\geq 0\) be positive real numbers. For \(r\geq 0\) define the power mean \[ M_r(x,y;\lambda)= (\lambda x^r+ (1-\lambda)y^r)^{1/r} (r\neq 0);\;= x^\lambda y^{1-\lambda} (r= 0), \] where \(\lambda\in [0,1]\). A positive function \(f\) is called \(r\)-convex, if for all \(x,y\in [a,b]\) and \(\lambda\in [0,1]\) we have \[ f(\lambda x+(1- \lambda)y)\leq M_r(f(x), f(y);\lambda). \] Let \(E\) be the Stolarsky mean of parameters \(r\), \(s\). The authors prove the following important theorem: If \(f\) is \(r\)-convex, then \(M_p(f)\leq E(f(a),f(b); r,p+r)\) for all \(p\geq 0\). This generalizes one of the Hadamard's inequalities, namely \[ {1\over b-a} \int^b_a f(t)dt\leq (f(a)+ f(b))/2. \] Another result extends an inequality by Fink, Mond and Pečarić, and is based on the Fejér generalization from 1906 of the Hadamard inequalities.