Two inequalities for convex functions

From MaRDI portal





The origin of the inequalities studied in this short note is a conjecture in Additive Number Theory, made by P. Erdős: with the sequence \(S\) of positive integers \(a_0<a_1<\cdots<a_n\) such that the additive sum \(\sum(S)=\sum_0^n\delta_ia_i\) contains only distinct numbers, where \(\delta_i=0,1\), one would have \[ \sum_0^n\frac{1}{a_i}\leq\sum_0^n\frac{1}{2^i}. \] Though one can easily verify with some examples that the inequality is the best upper bound, the condition in the conjecture is indeed a very strong one, and the conjectured inequality probably is not the only parameter to capture its real essence in terms of the additive structure of the sequence. It turned out that the conjecture was proved by P. Frenkel in 1998 via a purely analytic inequality: \[ \sum_0^nf(a_i)\leq\sum_0^nf(2^i), \] for any convex decreasing function. This result was further extended to \[ \sum_0^nf(\alpha_i)\leq\sum_0^nf(\beta_i),\quad\text{ if}~\alpha_1<\cdots<\alpha_n~\text{ and}~ \sum_0^k\alpha_i\geq\sum_0^k\beta_i,~\forall k, \] by Y. Chen in 2000. The results in this note generalizes Chen's results further. It shows first a result with non negative coefficients \(k_i\). That is \[ \sum_0^nk_if(\alpha_i)\leq\sum_0^nk_if(\beta_i),\quad\text{ if}\quad \sum_0^kk_i\alpha_i\geq\sum_0^kk_i\beta_i,~\forall k. \] Based on this it then shows \[ \sum_0^nf(\alpha_i)g(\alpha'_i)\leq\sum_0^nf(\beta_i)g(\beta'_i),\quad\text{ if}\quad \sum_0^k\alpha_i\geq\sum_0^k\beta_i~\text{ and}~\sum_0^k\alpha'_i\geq\sum_0^k\beta'_i,~\forall k, \] where \(\alpha'_1<\cdots<\alpha'_n\) and \(g\) is also a non-negative convex decreasing function.











This page was built for publication: Two inequalities for convex functions

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2577874)