Does any one know any bounded integral operator?

To put it in another way, I am given the following equation

$$ \int_{a}^{b} f(x,y) X(y) dy = b(x) $$

where $f(x,y)$ and $b(x)$ are known.

  • What conditions do I need in order to have a unique solution for $X(y)$?
  • How can I find $X(x)$?

Thank you

  • $\begingroup$ I suppose complete uniqueness is out of the question: Let $X$ be a solution, $X'$ a function that differs from $X$ only in a zero set of points. Then $X'$ is also a solution. For every reasonable integral I can think of, finite sets will have measure zero. Hence, the result will not be unique, since you can change every solution in a finite numbers of points and get another solution. $\endgroup$ – Johannes Kloos Mar 10 '12 at 15:36
  • $\begingroup$ Yeah, you are right. Let's put it in this way: Give one solution which holds for any arbitrary f(.,.) and b(.)? What condition(s) are needed to guarantee the uniqueness? $\endgroup$ – Daniel Oct 31 '12 at 15:44
  • $\begingroup$ Do you want absolute uniqueness or uniqueness up to some equivalence relation (e.g., equivalent almost everywhere)? $\endgroup$ – Johannes Kloos Oct 31 '12 at 15:55
  • $\begingroup$ Any kind of information could be useful. Do you have any ideas about such equivalence relation? $\endgroup$ – Daniel Oct 31 '12 at 15:59
  • $\begingroup$ Try "equivalent almost everywhere": Two functions $f, g$ are equivalent almost everywhere if the set $\{x | f(x) \neq g(x) \}$ has measure zero. This is still a very weak condition here, for example: Choose $f(x,y) = 1$ and $b(x) = 0$. Then you want to solve the equation $\int_a^b X(y) \operatorname{d}y = 0$, which does have infinitely many solutions, e.g., all functions of the form $X(y) = c \cdot \left(y - \frac{a+b}{2}\right)$. I am no expert in higher analysis, but I doubt there is any integral operator that will give you unique solutions except when adding some other stronger constraints. $\endgroup$ – Johannes Kloos Oct 31 '12 at 16:08

I found this useful1 : http://en.wikipedia.org/wiki/Fredholm_integral_equation

Using the mentioned Fourier transform formula, $ X(y) = \mathcal{F}_\omega^{-1}\left[ {\mathcal{F}_x[b(x)](\omega)\over \mathcal{F}_x[f(x,y)](\omega)} \right]=\int_{-\infty}^\infty {\mathcal{F}_x[b(x)](\omega)\over \mathcal{F}_x[f(x,y)](\omega)}e^{2\pi i \omega x} \mathrm{d}\omega $

it should be Lebesgue measurable:

$ \int_{-\infty}^\infty |b(x)| \, dx < \infty $

$ \int_{-\infty}^\infty |f(x,y)| \, dx < \infty $

One necessary condition is $\mathcal{F}_x[f(x,y)](\omega) \neq 0$. The problem with this method is that Fourier transform is not easy to calculate, though I can use DFT.

  • $\begingroup$ You can do this only if $f(x,y)$ is of the form $f(x-y)$ which is known as the difference kernel. $\endgroup$ – Mhenni Benghorbal Feb 19 '13 at 19:33
  • $\begingroup$ Oh, I see the problem with that .... $\endgroup$ – Daniel Apr 26 '13 at 20:21

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.