0
$\begingroup$

If I've got $f: x \to f(x)$, one may define the Arithmetic Dissidence $\delta_A[f(x)]$ as the real value of the difference between the length following the curve of $f$ and the length of the $x$ axis (we could define it on intervals or on the whole axis). You may also want to define the Geometric Dissidence $\delta_G[f(x)]$ as the ratio of these lengths.

For example, if can be easily shown that for a function $f: x \to |x|$, $\delta_G[f(x)]=\sqrt2$.
For a distribution $\mathcal{T}(1-|x|)_{[-1, 1]}$, $\delta_G[\mathcal{T}]$ can't be different from $1$ but $\delta_A[\mathcal{T}]=2(\sqrt2-1)$.

I need to find the value for $\delta_A[f(x)]$ if it's finite for $f=\lambda e^{\frac{-x^2}{2\sigma^2}}$. Otherwise, I'd need the value of $\delta_G[f(x)]$. I've still found no way of finding these values by integrating $d\ell=\sqrt{dx^2+df^2}$ but maybe you know one way.

Thanks for your help,

$\endgroup$
2
  • $\begingroup$ In what form would you like the answer? (Also, what do you mean by "$\delta_G[\mathcal T]$ can't be different from $1$"?) $\endgroup$
    – LSpice
    Commented May 31, 2021 at 3:24
  • $\begingroup$ I would like some expansion for the values of the Dissidences. (I meant that, in this case, it was $1$ : since I was talking about Dissidence over the whole real set and not only over a finite interval (where it could have been larger than $1$), $\lim_{x\to\infty}\frac{x+2(\sqrt{2}-1)}{x}=1$). $\endgroup$ Commented May 31, 2021 at 16:43

1 Answer 1

2
$\begingroup$

You seek the quantity $$\delta_G(x,\lambda)=\frac{1}{x}\int_0^x \sqrt{1+f'(y)^2}\,dy,\;\;\text{with}\;\;f(x)=\lambda e^{-x^2/2},$$ where I have rescaled $x\mapsto x/\sigma$ to remove the parameter $\sigma$. There is no closed form answer for this integral, but for large $x$ it decays as $$\delta_G(x,\lambda)\approx 1+\lambda^2\frac{\sqrt{\pi}}{8x},\;\;x\gg 1,\;\;\lambda\ll 1.$$ For small $x$ it grows as $$\delta_G(x,\lambda)\approx 1+\lambda^2\frac{x^2}{6},\;\;x\ll 1.$$ Here is a plot of $\delta_G(x,\lambda)$ for $\lambda=1$ (green) and the small-$x$ and large-$x$ asymptotes (blue and orange).


The alternative quantity $$\delta_A(x,\lambda)=\int_0^x \sqrt{1+f'(y)^2}\,dy-x$$ approaches a constant in the limit $x\rightarrow\infty$. For $\lambda\ll 1$ I find $$\lim_{x\rightarrow\infty}\delta_A(x,\lambda)=\frac{1}{8}\lambda^2\sqrt{\pi}.$$ Here is a plot of this large-$x$ limit as a function of $\lambda$. The orange curve is the exact result, the blue curve the small-$\lambda$ approximation.

$\endgroup$
4
  • $\begingroup$ Thank you for this quick answer ! I did get myself to the point I couldn't find a closed form for this integral. I'm interested in the value for the whole curve so for $x\to\infty$. $\endgroup$ Commented May 31, 2021 at 16:30
  • $\begingroup$ For sure $\delta_A$ is positive but does it diverge ? If it diverges I don't understand how $\delta_G$ could approach $1$ as $x$ gets infinitely large. If $\delta_A$ doesn't diverge, I'd like to know its expansion and if $\delta_G$ is not $1$, I'd also like to get its expansion. $\endgroup$ Commented May 31, 2021 at 16:37
  • $\begingroup$ (but apparently $\delta_G$ is indeed $1$ which kind of makes me think that $\delta_A$ doesn't diverge) $\endgroup$ Commented May 31, 2021 at 16:47
  • $\begingroup$ I have added the results for $\delta_A$, including the large-$x$ limit. This is finite, for small $\lambda$ it grows as $\lambda^2$. $\endgroup$ Commented May 31, 2021 at 17:39

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .