This question was posted in https://math.stackexchange.com but I got hardly any view. If posting here is an objection please let me know I would delete it immediately.

This question has evolved from watching the video: https://www.youtube.com/watch?v=dh5hHpJ79jc

In the video the speaker talked about white noise case and $\alpha$ related to the Holder condition, (wikipedia link to Holder condition https://en.wikipedia.org/wiki/H%C3%B6lder_condition)

The question: Is there a relation between SNR (https://en.wikipedia.org/wiki/Signal-to-noise_ratio) and Holder condition?

Let me be more specific, As I understand, Rough path theory (https://en.wikipedia.org/wiki/Rough_path) is related to the solution of differential equation driven by non-smooth signals and non-smoothness is defined in terms of Holder condition. In the video it clearly relates the white noise case to the value of $\alpha$. Consider our driver is simple sine wave, which is very smooth, but if we add a noise then we can define SNR of the driver based on the strength of the noise and at the same time can we define $\alpha$ based on the strength on the noise?

In some sense I am trying to define rough path theory in terms of SNR, is it possible? Any comments would be highly appreciated

  • $\begingroup$ Hölder exponent $h$, Hurst exponent $H$, and fractal dimension $D$ (with $D=2-H=1-h$) describe the correlations in the fluctuations of the signal, not the strength of the fluctuations as quantified by the SNR. $\endgroup$ Jul 10, 2018 at 8:16
  • $\begingroup$ @CarloBeenakker Thank you for your comments, but I did not understand how it is relates to the correlations. Mmay I request few more explanation or references which would help me to understand. $\endgroup$
    – Creator
    Jul 10, 2018 at 8:32

1 Answer 1


The Hölder exponent $h$ and the Hurst exponent $H$ (related by $h=H-1$ for a continuous signal) characterise correlations in the noise, not the size of the noise (as quantified by the signal-to-noise ratio). So these are different quantities. This plot illustrates two signals with the same signal-to-noise ratio but different $H$ (short-term correlations on the left, long-term correlations on the right).

source: On the relationship between the Hurst exponent, the ratio of the mean square successive difference to the variance, and the number of turning points

  • $\begingroup$ Please excuse my ignorance, I can understand Hölder exponent h (or alpha in the question) is greater then 0 for continuous signal, but I still could not get the relation h=H-1. In the paper, I did not get any mention about Hölder exponent h. Am I missing something obvious? $\endgroup$
    – Creator
    Jul 10, 2018 at 19:39
  • $\begingroup$ the relation between Hölder and Hurst exponents is discussed, for example, here arxiv.org/abs/cond-mat/0208028 (equation 2 and following text) $\endgroup$ Jul 10, 2018 at 20:04
  • $\begingroup$ May I ask how do you know the plots have same signal to noise ratio? $\endgroup$
    – Creator
    Jul 19, 2018 at 3:04
  • $\begingroup$ just visual inspection: the curve $x(t)$ has variance close to unity in both plots. $\endgroup$ Jul 19, 2018 at 6:29
  • $\begingroup$ en.wikipedia.org/wiki/Fractional_Brownian_motion in the link the difference of H clearly shows that (visually) that SNR varies. May I please request your view? $\endgroup$
    – Creator
    Jul 19, 2018 at 7:25

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.