A complementary notebook on photoz in scikit-learn context is available for a PR

I am calling for volunteers to review my new notebook which is complementary to Melissa’s notebook on PhotoZ. Then I should assign the PR on github in order to merge it on main branch
Many thanks.


1 Like

Reviewed and approved! Thanks Sylvie.

Hello Melissa,

Now I have merged my two notebooks on PhotoZ with scikit learn.

  • The first one is a kind of introduction and a brute comparison.
  • The second one is on optimisation of hyper-parameters.

Both are ready for a live demo. To speed up the demo, I have screened the PhotoDetection part. The part one could be shown during the first part of the delegate assembly. The second part could be shown during the second part of the delegate assembly in a dedicated PhotoZ session for those who want to to deeper with ML and scikit-learn.

Note it is not necessary to run the whole notebooks entirely. For example, for the part 2, it is possible to focus on the linear model (Ridge) optimisation only. It would depend on the time at our disposal.
Best regards.


Hi Sylvie,

I approved the review you requested on PR#18, the updates to the photoz/README.md file.

Would you like to present this Friday in the Delegate Assembly? There are, so far, no other presenters scheduled for this Friday. The next Delegate Assembly after this week would be April 8.


Hi Melissa,

Yes I accept to present my notebook. Please remind me where is the slide desk to write a few preliminary instructions.

Best regards.

Awesome! You’re our presenter for this Friday March 25 (assemblies schedule).

Here’s an editing link to the slide deck for Fri Mar 25:

Thanks again!

Hi Sylvie, thank you for providing this excellent learning tool, and for documenting it so well! I’d like to know more about where the expected detected magnitude contributions calculation comes from, i.e.
# some parameters for prob(imagnitude)
alpha = 15.0
beta = 2

# prob(imagnitude) distribution
p_imag = imag_grid**alpha*np.exp(-(imag_grid/(maglim-off))**beta)
p_imag /= p_imag.sum()
return p_imag

I looked at the reference from ixkael,
(GitHub - ixkael/Photoz-tools)/**Photoz galaxy survey mock and N(z) inference.ipynb

but I couldn’t find a reference for their calculation. You provide what I think is the appropriate (2016) reference:
Hierarchical Bayesian inference of galaxy redshift distributions from photometric surveys Boris Leistedt, Daniel J. Mortlock and Hiranya V. Peiris (page 4)

and I found a similar looking equation for p(z|t,m), but it’s not the same.

Thank you again for this code,