I am having trouble understanding this definition, [Arnold: Random Dynamical Systems]:

Definition: A measurable random dynamical system on the measurable space $(X,\mathcal{B})$ over a metric dynamical system $(\Omega,\mathcal{F},\mathbb{P},(\theta(t))_{t\in\mathbb{T}})$ as a map $\phi$ with time $\mathbb{T}$ is a mapping

$$\phi:\mathbb{T}\times\Omega\times X\to X,$$

where $\phi$ is measurable and satisfies the cocycle property. $\blacksquare$

I understand that a metric dynamical system is a family of iterative mappings on the space $(\Omega,\mathcal{F},\mathbb{P})$ that act as endomorphisms, but I don't understand how it all ties together. As far as I can tell, $\phi$ is a deterministic map.

I don't understand how randomness is described. Can anyone enlighten / inspire me?


For every $(t,\omega,x)$ in $\mathbb T\times\Omega\times X$, $\phi(t,\omega,x)$ is some point in $X$ hence, for every $(t,x)$ in $\mathbb T\times X$, $\phi(t,x):\omega\mapsto\phi(t,\omega,x)$ is a random variable with values in $X$.

Furthermore, one asks that $\phi(t+s,x)=\phi(t,\phi(s,x))$ for every $(t,s,x)$ in $\mathbb T\times\mathbb T\times X$, in the sense that $\phi(t+s,\omega,x)=\phi(t,\omega,\phi(s,\omega,x))$ for every $(t,s,\omega,x)$ in $\mathbb T\times\mathbb T\times\Omega\times X$. In other words, each $\phi(\ ,\omega,\ )$ is a flow on $\mathbb T\times X$.

Edit: Here is a dynamical system. Assume that $\mathbb T=\mathbb N_0$, $X=\mathbb R$, that $\Omega=\{0,1\}^\mathbb N$ is endowed with its cylindrical sigma-algebra, and consider the shifts $(\theta_n)_{n\geqslant0}$ defined by $\theta_n(\omega)=(\omega_{k+n})_{k\geqslant1}$ for every $n\geqslant0$ and $\omega=(\omega_k)_{k\geqslant1}$ in $\Omega$, for example, $\theta_6(\omega_1,\omega_2,\omega_3,\ldots)=(\omega_7,\omega_8,\omega_9,\ldots)$. Finally, define $\phi(1,\omega,x)=A_{\omega_1}(x)$ for every $x$ in $X$ and every $\omega=(\omega_k)_{k\geqslant1}$ in $\Omega$, where $A_0$ and $A_1$ are some given functions on $\mathbb R$. In words, $\phi(1,\ ,\ )$ applies $A_0$ or $A_1$ to $x$, the index $0$ or $1$ being chosen uniformly randomly.

Let me suggest that you check that, for every $n\geqslant0$, $$ \phi(n,\omega,x)=A_{\omega_n}\circ \cdots\circ A_{\omega_2}\circ A_{\omega_1}(x), $$ and that the cocycle property holds in this setting.

| cite | improve this answer | |
  • $\begingroup$ Thanks, yeah this makes sense, however I'm still confused by the cocycle property. $\phi(t+s,\omega ) = \phi(t,\theta(s)\omega ) \circ \phi(s,\omega ).$ What does the movement of $\omega$ through the metric space imply / represent? $\endgroup$ – David Simmons Oct 17 '13 at 22:52
  • $\begingroup$ Sorry but $\omega$ moves through the probability space. The alea $\omega$ is applied to the particle starting at time $0$. The alea $\theta_s(\omega)$ is what starts to be applied to the particle at time $s$. $\endgroup$ – Did Oct 18 '13 at 5:34
  • $\begingroup$ I guess this is what ultimately confuses me, why do we plot the trajectory of $\omega$ through $\Omega$? I mean, what does $\phi(s,\omega)$ represent. \\ Presumably it's the statement "given $\omega$ has occurred, $x$ will be sent to $\phi(s,\omega)x$ after time $s$ has elapsed" \\ But if the previous statement is true, doesn't that imply (from the cocycle property) that if $\omega$ has occured at time $0$ then we know $\theta(s)\omega$ will occur after time $s$? Do you understand my confusion? $\endgroup$ – David Simmons Oct 18 '13 at 9:47
  • $\begingroup$ See Edit. $ $ $ $ $\endgroup$ – Did Oct 18 '13 at 14:56
  • $\begingroup$ Fantastic! Thank you ever so much $\endgroup$ – David Simmons Oct 18 '13 at 18:08

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.