Negation of a property, that hold almost everywhere

134 Views Asked by At

Let $1\leq p\leq \infty$, $I\subseteq\mathbb R$ an open interval and $f\in L^p(I)$ and $\Phi\colon \mathbb R\to\{\text{true},\text{false}\}$ be a predicate.

Sometimes it is written something like that: $$ \Phi(f(x))\qquad\text{almost everywhere in $I$}.\tag{P} $$


I think this means the following: $$ \exists \tilde f\in f\colon\quad \lambda(\{ x\mid \neg \Phi(\tilde f(x)\}) = 0 \tag{P1}$$, where $\tilde f$ is a representant of $f$ and $\lambda$ is the Lebesgue measure. This means its negation would be $$ \forall \tilde f\in f\colon\quad \lambda(\{ x\mid \neg \Phi(\tilde f(x)\}) \neq 0 \tag{N1}$$


Another way of putting (P) would be $$ \exists \tilde f\in f\colon\quad \exists M\in N(I)\colon\quad\forall x\in I\setminus M\colon\quad \Phi(\tilde f(x))\tag{P2}$$ where $N(I)$ is the set of all Lebesgue measureable subsets of $N$ with vanishing Lebesgue measure. Then its negation would be $$ \forall \tilde f\in f\colon\quad \forall M\in N(I)\colon\quad\exists x\in I\setminus M\colon\quad \neg \Phi(\tilde f(x))\tag{N2}$$


My question are:

  1. Are (P1) and (P2) correct interpretations of (P)?
  2. If so, are they equal?
  3. Are (N1) and (N2) correct negations of (P1) and (P2)? (I'm pretty sure about that)
  4. If so, how can they be interpreted
  5. Last but not least: How does the negation of (P) correspond to $$\neg \Phi(f(x))\qquad\text{almost everywhere in $J\subseteq I$, where } \lambda(J)\neq 0. \tag{N}$$
1

There are 1 best solutions below

0
On

Firstly, your $p$ is irrelevant. We may as well just be talking about measurable functions with properties that are true a.e.

I would say that equivalence of (P1) and (P2) is pretty straightforward. Assume one, and try to prove the other, and vice versa. However, I would go further and say that in fact, both of your (P1) and (P2) are a bit weaker than the truth, in that if they hold for any representant, they hold for all. This is because any two representants differ only on a set of measure $0$. But regardless, they are equivalent.

I think (N1) is a little wrong. In particular there are no guarantees that the set where the proposition is false is measurable. You could say that for all representants $\tilde f$ and $A\in\sigma(\mathbb{R})$ with $\{x | \neg\Phi(\tilde f(x))\} \subseteq A$, $\lambda(A) > 0$. This is much more obviously equivalent to your second condition.

(N) is a stronger condition than not (P). Its saying that the set where the proposition fails contains a non null set, whereas the correct interpretation of not (P) is that the set where the proposition fails cannot be contained in a null set.

Hope there's some clarity in that for you.