yonilc / ddecc Goto Github PK

Hi,

I may sound lame from the question I asked but please allow me to give a context. So, I am not from communications background. Just a pure Computer Vision researcher. I have recently come across a problem where I need to transfer random bit string over a network and recover the same with zero loss. I experimented with a linear autoencoder architecture with one linear layer each and of course I could recover the string with almost zero loss. But the problems can't be so simple, right? Now when I tried to add noise to my data after encoding step; my decoder isn't able to recover the string due to obvious non-linearity induced in between the network.

Then I read more and came across communications field and could relate this with my problem statement because that's what I am trying to do -> just transmitting information bits and trying to recover it with ECC embedded.

I came across your work and trying to understand if this is it what I am looking for. But when I checked your dataset (POLAR); I found they're some kind of vector matrices with ones and zeros extending upto a length like [000011111110000], meaning no mixed bits like [01010000111010100110011].

So; please help me understand if such random bit strings are possible to recover and whether I am on the right path?

Thanks a lot for this work!