Let $\mathcal{F}$ be the set of continuous and strictly increasing functions from $[0,1]$ to $[0,1]$ with $f(0)=0$. Is there a closed form for $$\inf_{f\in\mathcal{F}}\sup_{0\le x\le 1}\frac{(1-x)f(x)}{\int_0^1f(t)\,\mathrm{d}t}?$$

That is, the ratio of the following dark area to the integral area.

Note: this question is similar to this one, but this time I want to minimize the ratio when $f$ varies.

  • $\begingroup$ How did you obtain $\frac{1}{4(1-\varepsilon)}$? $\endgroup$ – Uskebasi Apr 12 at 14:26
  • $\begingroup$ @Uskebasi By symmetry we can assume $x\le 1-\epsilon$, then $f(x)=\frac{x\epsilon}{1-\epsilon}$, so we have $\frac{(1-x)f(x)}{\int_0^1f(t)\,\mathrm{d}t}=\frac{(1-x)x\epsilon}{(1-\epsilon)\epsilon}$, which has a maximum of $\frac{1}{4(1-\epsilon)}$. $\endgroup$ – xskxzr Apr 12 at 15:16
  • $\begingroup$ @xskxzr Where does this nice problem come from? $\endgroup$ – Robert Z Apr 16 at 11:32
  • $\begingroup$ @RobertZ It comes from a research, and I formalize it as this mathematical problem by myself. $\endgroup$ – xskxzr Apr 16 at 14:29

The desired infimum is zero. For $\epsilon\in (0,1/2)$, consider the strictly increasing continuous function $$f(x)=\begin{cases} \frac{\epsilon x}{1-x}& \text{if $x\in [0,1-\epsilon]$},\\ x& \text{if $x\in [1-\epsilon,1]$.} \end{cases}$$ Then $$\int_0^1f(t)\,dt =\frac{\epsilon^2}{2}-\epsilon\log(\epsilon)$$ and $\sup_{x\in [0,1]}(1-x)f(x)=\epsilon(1-\epsilon)$. Hence, as $\epsilon \to 0^+$ $$\frac{\sup_{x\in [0,1]}(1-x)f(x)}{\int_0^1f(t)\,dt }=\frac{\epsilon(1-\epsilon)}{\frac{\epsilon^2}{2}-\epsilon\log(\epsilon)}=\frac{1-\epsilon}{\frac{\epsilon}{2}+\log(1/\epsilon)}\to 0^+.$$


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.