Numerical verification of condition for approximately midconvex functions (Q443926)

Let \(V\) be a convex subset of a normed space \(X\) and \(\alpha :{\mathbb{R}}_+ \to {\mathbb{R}}_+\). A function \(f:V\to \mathbb{R}\) is said to be \(\alpha \)-midconvex if \[ f(\frac{x+y}{2})-\frac{f(x)+f(y)}{2}\leq \alpha (\|x-y\|),\quad x,y\in V. \] Every such a continuous function \(f\) satisfies the following estimation \[ f(tx+(1-t)y)-tf(x)-(1-t)f(y)\leq \sum _{k=0}^{\infty }\frac{1}{2^k }\alpha (d(2^{kt} \|x-y\|)),\quad x,y\in V, t\in [0, 1], \] where \(d(t):=2\)dist\((t, \mathbb{Z})\) for \(t\in [0, 1]\). In [\textit{J. Makó} and \textit{Z. Páles}, J. Math. Anal. Appl. 369, No. 2, 545--554 (2010; Zbl 1201.26003)] it was proved that this estimation is optimal for functions of type \(\alpha (r)=r^p\) for \(p\in (0, 1)\). In this paper, a different approach to the verification of the optimality in a class of continuous piecewise linear mappings is presented. To do this, the authors use a result from [\textit{J. Tabor, J. Tabor} and \textit{M. Żołdak}, Aequationes Math. 80, No. 1-2, 227--237 (2010; Zbl 1213.26015)] and an algorithm based on interval arithmetic.