2
$\begingroup$

I would like to use the dataset for this kaggle but the competition was private, is closed and I have been unable to contact anyone directly.

The challenge is machine learning for photo-z estimation from u,g,r,i,z magnitudes. Out of curiosity, I would like to apply an ML technique not apparent in code samples submitted to the competition.

Photo-z estimation is the process or method of estimating the redshift (z) of an object based solely on its brightness, or magnitude, through various colour filters (u, g, r, i, etc.). This is desirable because sky surveys typically collect all such magnitudes for distant* objects automatically; this is quick and efficient, whereas measuring the spectroscopically red-shift requires particular attention, i.e. more effort and more time. Consequently, photo-z estimation can provide more redshift values for a given investment of effort, which is important for generating observational statistics.

(* We are talking about cosmologically distant here, objects nearby, in our local group of galaxies for example, don't have a "cosmological redshift)

Does anyone know whether the data set is available elsewhere?

$\endgroup$
6
  • 1
    $\begingroup$ Could you maybe elaborate a bit in your question on what the problem is about, please? I am indeed able to read what the competition is about, but it would be great to have a high-level summary in the question as well. $\endgroup$
    – B--rian
    May 9 at 20:41
  • 1
    $\begingroup$ @B--rian I have added the requested elaboration on why photo-z estimation is of interest. HTH. $\endgroup$ May 10 at 6:42
  • $\begingroup$ I found kids.strw.leidenuniv.nl/DR3/ml-photoz.php which seems to be a different dataset, but maybe helpful nevertheless. $\endgroup$
    – B--rian
    May 11 at 20:02
  • 1
    $\begingroup$ @B--rian Thanks. I will take a look at that. I'm trying to avoid having to set up my own query on e.g. SDSS ;) Because I'm not a astronomer it's much easier if the selection and features are just given. I would still like that specific dataset to compare with the competitors' results though. $\endgroup$ May 12 at 12:39
  • $\begingroup$ @B--rian the KiDS datasets seem to be dataset of objects with photometrically estimated red-shifts, i.e. the outputs (2 methods) rather than the raw data - unless you know specifically otherwise. So, unfortunately not useful to me. $\endgroup$ May 12 at 20:04

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.