2
$\begingroup$

I have $N$ mixtures consisting of one non-Gaussian source (I know the distribution) and many (more than $N$) Gaussian sources.

I also know how my non-Gaussian source is mixed into the signal (I know the elements of A that relate my non-Gaussian source to the mixture in $x=As$).

I have no information about my Gaussian sources, except for the fact that they are Gaussian and separated in space. My Gaussian sources have an unknown mean that is not zero.

Is there a way I can reconstruct my non-Gaussian source? I tried it with one Gaussian source and it works. But as soon as I activate multiple Gaussian sources it degrades my reconstruction too much.

Any literature hints would be appreciated!

$\endgroup$
0
$\begingroup$

In ICA, there can be at most one Gaussian source. Check Hyvärinen's papers and ICA book.

$\endgroup$
2
  • 1
    $\begingroup$ That would be more helpful to the questioner if you gave us a reference to some of the papers and perhaps the publication details of the book. $\endgroup$ – mdewey Sep 6 '16 at 16:16
  • $\begingroup$ This is being automatically flagged as low quality, probably because it is so short. At present it is more of a comment than an answer by our standards. Can you expand on it? We can also turn it into a comment. $\endgroup$ – gung - Reinstate Monica Sep 6 '16 at 17:20

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.