How do I prove the following:

Let $f$ be a measurable function on $[0,1]$ such that $f$ is finite almost everywhere. Then for any $\varepsilon \gt 0$, $\exists$ a bounded measurable function $g$ such that $$ \mu\{x\in [0,1]: f(x)\neq g(x)\}\lt \varepsilon .$$ $\mu$ is the Lebesgue measure.

  • 1
    $\begingroup$ To make the question better, please indicate where you ran into the problem - what context did you see it in, what book are you using, etc. This will help others answer your question in a way that is more useful to you. $\endgroup$ – Carl Mummert Nov 23 '11 at 15:15
  • $\begingroup$ As a hint, consider for each $n \in \mathbb{N}$ the set $A_n = \{ x : |f(x)| \in [n,n+1)\}$. $\endgroup$ – Carl Mummert Nov 23 '11 at 15:16
  • $\begingroup$ @CarlMummert: I'm studying for an exam on Lebesgue measure. I saw the question in one of the old exams. thanks for the hint. $\endgroup$ – AKM Nov 23 '11 at 16:04

For each positive integer $n$, let $A_n=\{x\in[0,1] : |f(x)|\le n\}$. Note that $A=\bigcup\limits_{n=1}^\infty A_n$ is the set of points where $f$ is finite.

Now, each $A_n$ is measurable and $A_n\subset A_{n+1}$ for each $n$; thus, $$ \lim_{n\rightarrow\infty} \mu(A_n) =\mu (A)=1. $$

So, given $\epsilon>0$, we may (and do) choose $N$ with $\mu( A_N^C\,)\lt\epsilon$.

Define $g(x)=\cases{f(x),& x\in A_N\cr 0, &x\in A_N^C}\ $.

Then $g$ is bounded, measurable (since the $A_n$ are), and $$\mu\bigl(\{x\in[0,1]:f(x)\ne g(x)\} \, \bigr)=\mu(A_N^C\thinspace)\lt\epsilon. $$

| cite | improve this answer | |
  • $\begingroup$ Thanks for your answer. Is there a particular reason, why you defined $g$ the way you did? Also why is $ \mu\bigl(\{x\in[0,1]:f(x)\ne g(x)\} \, \bigr)=\mu(A_N^C\thinspace) $ $\endgroup$ – AKM Nov 23 '11 at 16:05
  • 1
    $\begingroup$ Just look at the definition: $g$ is just the function $f$ with "the big part lopped off". The basic problem is to show that the measure of the set where $f$ is big ($A_n^C$) tends to zero. $\endgroup$ – David Mitra Nov 23 '11 at 16:16

This is a slight elaboration on David Mitra's answer, but is too long to fit in a comment box.

To understand this question and how to solve it, you have to ask yourself what it means for the measure of the set where $f(x) \neq g(x)$ to be small (i.e. $< \epsilon$).

The smallest possible set is the empty set; to say that the set where $f(x) \neq g(x)$ is empty is to say that $f$ and $g$ are equal.

Now we can't necessarily take $f$ and $g$ to be equal, because $f$ may not be bounded, but we want $g$ to be bounded.

So instead, we are going to allow $g$ to differ from $f$, but only on a small set (one of measure $< \epsilon$).

Intuitively speaking, the "closer" $g$ is to the original function $f$, the smaller the set on which they differ will be.

So we want to find a way to change $f$ as little as possible while making it bounded. How can we do this?

Well, just choose a threshhold $n$, and define $f(x) = g(x)$ if $f(x) \leq n$, and $g(x) = 0$ otherwise. (The particular choice of $0$ here is not important, any number of absolute value $\leq n$ would do.)

Now $g(x)$ and $f(x)$ coincide unless $|f(x)|$ is too big.

So the problem is reduced to showing that the set of points where $|f(x)|$ is too big (i.e. where $|f(x)| > n$), which is the set of points where $f$ and $g$ differ, is small.

Now you have to use something: after all, if instead of a function on $[0,1]$, we had $f(x) = x$ on the whole real line $\mathbb R$, then the set of points where $|f(x)| > n$ (which is now just the set $(-\infty,-n) \cup (n,\infty)$) has infinite measure, and so is not small at all.

This is what David Mitra shows in his answer: because the total measure of $[0,1]$ is finite, and because $f(x)$ takes finite values almost everywhere, the set of points whre $|f(x)|> n$ has arbitrarily small measure, if we take $n$ large enough. QED

Some final remarks: from your comments on David Mitra's answer, I get the impression that you are thinking about this question in a very formal way. I would recommend that you practice translating formulaic expressions into more intuitive terms, i.e. try to read $\{x \, | \, f(x) \neq g(x)\}$ as "the set where $f$ and $g$ differ", and try to read $\mu(X) < \epsilon$ as "the set $X$ is small". Then you will have more chance of understanding what is really involved in a question, and hence have a better chance of answering it.

| cite | improve this answer | |

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.