2
$\begingroup$

[I realize that this question may be too basic and likely to contain a fundamental misunderstanding, but similar questions such as this one on the site are beyond my math level. So please assume very limited understanding of measure theory. If at all possible an answer in plain English would be ideal.]


A probability space is defined as the triple ($\Omega,\mathcal{F},P$).

The sample space and the sigma-algebra of events of a probability space -($\Omega,\mathcal{F}$)- would seem to fulfill the conditions of a topological space, ($X,\tau$):

  1. There is always a null event ($P(\{\emptyset\} = 0)$) and a certain event ($P(\{\Omega\})=1$). This corresponds to the condition, "The empty set and $X$ itself belong to $\tau$" in a topology sapce.
  2. Any union of events belongs to $\mathcal{F}$ (closed under countable unions). The counterpart in topology spaces: "Any (finite or infinite) union of members of $\tau$ still belongs to $\tau$."
  3. The intersection of any number of events belongs to $\mathcal{F}$ (closed under countable intersections). In topology spaces: "The intersection of any finite number of members of $\tau$ still belongs to $\tau$."

Can we then say that the sample space $\Omega$ and the $\sigma$-algebra, ($\Omega, \mathcal{F}$), form a topology on $\Omega$?

I believe that ($\Omega, \mathcal{F}$) also fulfills the conditions of a measurable space, so I may be indirectly asking the difference between a topological and a measurable space.

Is the coincidence of criteria an accident among conceptually disparate mathematical objects: probability (or measurable) spaces v. topology spaces?

Is it just a matter of the geometric component of topology?

Intuitively, I see that when applying a function to these spaces, in the case of topology we may want to relate somehow subsets by proximity; and I can see how in assigning a probability measure in [0,1] to different events, the result will be similar the more related (closer?) these events are...


Post-mortem notes:

  • In a $\sigma$-algebra any countable union of elements in $\mathcal{F}$ have to be contained in $\mathcal{F}$. On the other hand for a topological space it is any arbritrary union (finite or infinite) of the elments of the topology $\tau$ has to be contained in $\tau$. $\sim \text{Comment 1 below.}$
  • By De Morgan's law and closure under complements of probability spaces, $\mathcal{F}$ is similary closed under countable intersections. Conversely, in topology it is the intersection of finite elements of $\tau$ that remains contained in $\tau$.
$\endgroup$
  • 5
    $\begingroup$ "Any union of events belongs to F" Nope, only countable ones. This is the reason why not every topology is a sigma-algebra. $\endgroup$ – Did Mar 29 '16 at 21:49
  • 1
    $\begingroup$ An example of what goes wrong: the sigma-algebra of Borel sets (or even Lebesgue-measurable sets) on $\Bbb R$ contains all the singletons $\{x\}$, $x\in\Bbb R$. The only topology on $\Bbb R$ (or any set) that contains the singletons is the discrete one. $\endgroup$ – user228113 Mar 29 '16 at 21:56
  • $\begingroup$ @Did Well, I had a feeling that it wouldn't be correct... No Google returns is always an ominous sign... Your point seems to offer a single differentiating feature, and it'd be great if you could unfold it into an answer for dummies... $\endgroup$ – Antoni Parellada Mar 29 '16 at 22:01
  • 2
    $\begingroup$ Unfold into what? I feel mine and @G.Sassatelli's comment together pretty much sum up what there is to understand to the matter (whether for "dummies" or not). $\endgroup$ – Did Mar 29 '16 at 22:06
  • $\begingroup$ Also, $99\%$ of well-behaved topologies are not closed under complementation, while this is a crucial feature of sigma-algebra. To be fair, both in geometry and analysis we are particular fond of connected topological spaces, i.e. topological spaces $(X,\tau)$ where the only $A\subseteq X$ such that $A\in\tau\wedge X\setminus A\in\tau$ are $X$ and $\emptyset$. $\endgroup$ – user228113 Mar 30 '16 at 9:12
6
$\begingroup$

No, topologies and measurable spaces are not the same notions ; some topologies on a set are not measurable spaces and some measurable spaces on a set are not topologies.

A topology $T$ usually generates a measurable space $M$ called the Borel $\sigma$-algebra, which is the smallest $\sigma$-algebra containing all open sets, i.e. containing the topology. We have $T=M$ if and only if $T$ is a measurable space, i.e. if and only if $T$ is closed under countable intersections ; this can happen but there are lots of counter-examples (for instance, topological manifolds).

A measurable space $M$ is a topology if and only if it is closed under arbitrary unions, not just countable ones. For instance, if we take an uncountable set $X$ and let $Y \subseteq X$ satisfy $Y \in M$ if and only if $Y$ or $X \backslash Y$ is countable, this defines a $\sigma$-algebra $M$ on $X$. It is not a topology because $X$ admits a subset $Y$ which is uncountable with uncountable complement and we can write $Y = \bigcup_{y \in Y} \{y\}$ (a union of subsets in $M$ which does not belong to $M$).

Hope that helps,

$\endgroup$
  • $\begingroup$ It's great, but I wish you had (if possible) dumbed it down a bit, steering away from set-based definitions. $\endgroup$ – Antoni Parellada Mar 29 '16 at 22:03
  • 4
    $\begingroup$ @AntoniParellada You are asking a somewhat technical question about the distinction between two mathematical objects which are defined in terms of set operations. I'm afraid there isn't any "dumbing down," step #1 in math is always to learn the language :) $\endgroup$ – Math1000 Mar 29 '16 at 22:16
  • $\begingroup$ Of course... Sometimes people ask for a "motivating example" or an "intuition"... $\endgroup$ – Antoni Parellada Mar 29 '16 at 22:19
  • 1
    $\begingroup$ @Antoni Parellada : I assure you that your question is set-based. The amount of set-theory I used is not hard, I kept it to a minimum. For instance, the measurable space I introduced is the smallest one needed to introduce the counting measure ; i.e. the one whose value on a finite set is the number of its elements. Feel free to ask for me details but I assure you this is the answer you want ; it gives you the distinction between the axioms of a topological space and those of a measurable space. $\endgroup$ – Patrick Da Silva Mar 30 '16 at 9:28

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.