0
$\begingroup$

Consider the case where $X$ is a random variable defined on a space of functions say $C[0,1]$. Then each sample $X(\omega)$ is a path of stochastic process on $[0,1]$ as opposed to the usual case where $X(\omega)$ is simply a number or vector. Is it possible to have an equivalent of probability density function for such $X$ so that we can kind of measure probability of $X$ taking certain paths? For example, $\int_Gf_X(g)dg\in \mathbb{R}$ would be probability of $X$ realizes to one of the collection of paths $G$ where each $g \in G$ is a path on $[0,1]$.

$\endgroup$
0
$\begingroup$

Talking about densities requires you to fix a reference measure that your density is against. You implicitly do this when you write an integral against $dg$. The problem is that it's not obvious what you mean by $dg$.

In finite dimensions we have a somewhat canonical choice of reference measure, the Lebesgue measure. However, there is no good analogue of the Lebesgue measure in infinite dimensions. Indeed, we have

Lemma: Let $X$ be an infinite dimensional normed space and let $\mu$ be a translation invariant measure on the Borel $\sigma$-algebra of $X$. Then $\mu(B(0,1)) = \infty$.

Unfortunately, this means that densities become less useful in the infinite dimensional setting.

Of course, we could fix some measure on $C[0,1]$ and ask if there is a density against that measure but then it's not obvious what kind of measure you'd want to fix and why that would be useful here.

$\endgroup$

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.