I'm trying to understand the following conclusion from this paper (see below for the relevant paragraphs):

enter image description here

I'm not sure whether they really mean that it follows from the statements of Lemma 3.2 (since I absolutely don't see how this conclusion is possible) or from its proof (which I could imagine).

Taking a closer look to the proof of Lemma 3.2-ii, we see that the result actually is that for all $r\in[r_0,1)$ and for all $\alpha\in(0,1)$, there is a $K\ge0$ such that $$\operatorname E\left[\rho_r(X^x_1,X^y_1)\right]\le\alpha\rho_r(x,y)+K\;\;\;\text{for all }x,y\in E\tag1.$$ And, as always, this extends to bounds for $\operatorname E\left[\rho_r(X^x_n,X^y_n)\right]$ by the semigroup property.

In the proof the additive $K$ corresponds to a bound on an integral restricted to an open ball. Maybe the assumption $\rho(x,y)\ge K_\ast$ can be used to show this integral is $0$?

enter image description here enter image description here enter image description here enter image description here enter image description here


1 Answer 1


No, you just have to take $K_*$ such that $K \leq (\alpha_*-\alpha)\rho(x,y)$, and so for example $K_* = \frac{K}{\alpha_*-\alpha}$.

If I am not missing something, a good way of doing the proof could be:

Since $\alpha<1$, there exists $\alpha_*\in (\alpha,1) \subset (0,1)$. Now take $K_* = \frac{K}{\alpha_*-\alpha}$ and $\rho(x,y) ≥ K_*$. Then $$ \begin{align*} \alpha\,\rho(x,y) + K &= \alpha\,\rho(x,y) + (\alpha_*-\alpha)\,K_* \\ &≤ \alpha\,\rho(x,y) + (\alpha_*-\alpha)\,\rho(x,y) = \alpha_*\,\rho(x,y) \end{align*} $$

  • $\begingroup$ Thank you for your answer. A few remarks: It should be $K_\ast:=\frac K{\alpha_\ast-\alpha}$, since otherwise $K_\ast$ is negative. And the first inequality in your displaced equation is an equality. $\endgroup$
    – 0xbadf00d
    Jun 13, 2020 at 18:37
  • $\begingroup$ Yes, corrected ;) $\endgroup$
    – LL 3.14
    Jun 13, 2020 at 18:51
  • $\begingroup$ They proceed in the proof of Lemma 3.8 by showing a rather simple inequality: i.stack.imgur.com/hJRkW.png. It seems like they are claiming that $$1-\alpha_\ast+\alpha_\ast d(x,y)\le\frac{1+\alpha_\ast\beta K_\ast}{1+\beta K_\ast}d(x,y),$$ but I'm not able to see how they obtain this. They say that it follows from $d(x,y)\ge1+\beta K_\ast$ and so I've tried to add $d(x,y)-(1+\beta K_\ast)\ge0$ to the right-hand side. However, I wasn't able to arrange the terms so that they match the right-hand side. Do you've got an idea? $\endgroup$
    – 0xbadf00d
    Jun 14, 2020 at 6:15

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.