0
$\begingroup$

I need to construct a function $f:[0,1] \to \Bbb R$ such that $f$ is not a step function, but for any $\epsilon \in (0,1)$ the restriction of $f$ to $[\epsilon,1]$ is a step function.

So my function would be:

$f(x)=\begin{cases}x^2, \text{ for $\in[0,1-\epsilon]$}\\f_i(x) , \text{ for $x\in(p_{n-1},p_n) \subset P_{i} \subset (1-\epsilon,1]$}\end{cases}$

Where $P_i=\{p_{i_0},p_{i_1},...,p_{i_k}\}\in\{P_1,P_2,...,P_k\}$ is a partition compatible with $f_i(x)$ on the subset $(1-\epsilon,1]$

Would this make any sense? I want the partitions to basically be dependent on the choice of $\epsilon$ we make, so that's why we have a set of partitions

$\endgroup$
2
  • $\begingroup$ If you want "a function $f$, such that for any $\epsilon \in(0,1)$" you have some desired property, then your function $f$ should not depend on $\epsilon$. The way this is worded, you fix $f$, then are free to choose $\epsilon$. So $f$ should not change with $\epsilon$. Do you mean you want a family of functions $(f_\epsilon)_{\epsilon\in(0,1)}$? $\endgroup$
    – Dayton
    Apr 29, 2020 at 16:48
  • $\begingroup$ @Dayton actually yes, it's a function, not a family of functions, i got it wrong.. (so it's only one function) $\endgroup$
    – user634512
    Apr 29, 2020 at 16:50

1 Answer 1

0
$\begingroup$

Consider the function $$f(x) = \begin{cases} \chi({\sin(1/x)}), \ &x\in(0,1] \\ 0, \ &x = 0 \end{cases}$$ where $$\chi({x}) = \begin{cases} 1, \ &x\geq0 \\ 0, \ &x< 0\end{cases}$$ Then on any interval $[\epsilon,1]$ with $\epsilon\in(0,1)$, the restriction $f|_{[\epsilon,1]}$ is a finite sum of characteristic functions and hence a step function. But on $[0,1]$ the sum is infinite and is therefore not a step function.

$\endgroup$

You must log in to answer this question.