0
$\begingroup$

Is there a cousin of the Fourier transform which obeys the following property: if I rescale and translate a function, then its transformed form, at each frequency, only depends on the original transformed function at that frequency? In detail:

Given a function $f$, we can take its Fourier transform to get $\mathcal{F}(f)$:

$\mathcal{F}(f)(\omega)=\int e^{-2\pi ix \omega}f(x) dx$

If we translate $f$ by $\delta$ to get $g(x)=f(x-\delta)$, the Fourier transform of $g$ obeys $\mathcal{F}(g)(\omega)=e^{-2\pi i\delta \omega}\mathcal{F}(f)(\omega)$. In other words, $\mathcal{F}(g)(\omega)$ is a function of $\mathcal{F}(f)(\omega)$, $\delta$, and $\omega$.

If we scale $f$ by $a$ to get $g(x)=f(x/a)$, the Fourier transform of $g$ obeys $\mathcal{F}(g)(\omega)=\frac{1}{|a|}\mathcal{F}(f)(\omega/a)$. In other words, $\mathcal{F}(g)(\omega)$ is no longer a function of $\mathcal{F}(f)(\omega)$, but instead is a function of $\mathcal{F}(f)(\omega/a)$.

Does there exist a hypothetical new transformation $\mathcal{T}$, like the Fourier transform, but which obeys the following property: If $g(x)=f(ax-\delta)$, then $\mathcal{T}(g)(\omega)$ is a function only of $\mathcal{T}(f)(\omega)$, $\omega$, $a$, and $\delta$.

$\endgroup$
3
  • 2
    $\begingroup$ If you postulate that the transform is given by "integrating against a kernel" (a function of two variables...), you'll see that this is very difficult to arrange... $\endgroup$ Commented May 16, 2022 at 22:06
  • $\begingroup$ Thanks, I got very stuck on this when I made that assumption. Can that assumption be relaxed to solve this problem? $\endgroup$ Commented May 16, 2022 at 23:26
  • 2
    $\begingroup$ Well, in fact (google "Schwartz' Kernel Theorem") in nearly every situation operators can be described by kernels. Lots of technicalities, naturally. :) But, basically, any continuous linear map that makes sense as a map from test functions (a very tiny nice class of functions) to distributions (a pretty large class...) on $\mathbb R$ (and such) is given by a distribution on $\mathbb R^2$... which is a generalization of "kernel". $\endgroup$ Commented May 17, 2022 at 0:53

0

You must log in to answer this question.

Browse other questions tagged .