Let $(X,\mu,f)$ be a two-sided full shift system. Assume that there is $t \in \mathbb{N}$ such that for every $n \in \mathbb{N}$ and $x \in X$, we can define $T(x)=f^{n+m(x)}(x)$, where $m(x) \leq t; $ $m(x) \in \mathbb{N}$ and that depends on $x$. I also assume that $T$ is measure preserving.

It is well-known that $h_{\mu}(f^n)=nh_{\mu}(f)$, where $h_{\mu}$ is the measure-theoretic entropy.

Can we use the above fact to say $h_{\mu}(T) \leq (n+t)h_{\mu}(f)$?

Edit: As it was mentioned in comments, $T$ is not necessarily measure preserving. I add the assumption that $T$ is measure preserving.

  • $\begingroup$ To begin with, your new transformation need not be measure preserving. $\endgroup$
    – R W
    Oct 18, 2022 at 11:24
  • $\begingroup$ @RW Thanks for your comment. Even, isn't it a measure preserving when $(X, \mu, f)$ is a full shift? $\endgroup$
    – Adam
    Oct 18, 2022 at 11:37
  • $\begingroup$ What is $\mathbb{N}$? $\endgroup$
    – Ville Salo
    Oct 18, 2022 at 12:58
  • $\begingroup$ @VilleSalo Natural numbers $\endgroup$
    – Adam
    Oct 18, 2022 at 13:18
  • 1
    $\begingroup$ @IlkkaTörmä The reason that I asked questions is: I have a preserving function like $T$, where it is some iteration of $f$, and I want to see whether the above entropy relation holds or not. Unfortunately, I can not exactly write the function as it is so complicated, but I knew $f$, which is some iteration of $T$ is preserving. $\endgroup$
    – Adam
    Oct 18, 2022 at 15:59

1 Answer 1


If I understand right, your function $m$ (for a fixed $n$) takes on only finitely many values, which are all measurable sets. You can then define the partition $\{m^{-1}(i)\}$ and the associated finite $\sigma$-algebra $\mathcal{B}$.

If we define $\mathcal{A}$ to be any finite sub-$\sigma$-algebra of your original $\sigma$-algebra of measurable sets, then it's well-known (Corollary 4.10 of Walters) that $k^{-1} H_\mu\left(\bigvee_{i=0}^{k-1} T^{-i} \mathcal{A} \right)$ converges (in fact decreases) to $h_\mu(T, \mathcal{A})$.

In your case, I think by definition $\bigvee_{i = 0}^k T^{-i} \mathcal{A}$ is contained in $\bigvee_{i = 0}^{k(n+t)-1} f^{-i} (\mathcal{A} \vee \mathcal{B})$. Therefore, $k^{-1} H_\mu\left(\bigvee_{i=0}^{k-1} T^{-i} \mathcal{A} \right) \leq k^{-1} H_\mu\left(\bigvee_{i = 0}^{k(n+t)-1} f^{-i} (\mathcal{A} \vee \mathcal{B}) \right) = (n+t) (k(n+t))^{-1} H_\mu\left(\bigvee_{i = 0}^{k(n+t)-1} f^{-i} (\mathcal{A} \vee \mathcal{B}) \right).$

But by the above, the first quantity approaches $h_\mu(T, \mathcal{A})$ and the final quantity approaches $(n+t) h_\mu(f, \mathcal{A} \vee \mathcal{B})$. Therefore, $h_\mu(T, \mathcal{A}) \leq (n+t) h_\mu(f, \mathcal{A} \vee \mathcal{B})$, and then taking the supremum over $\mathcal{A}$ yields $h_\mu(T) \leq (n+t) h_\mu(f)$.


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.