QUBIC

Minutes:

• Haochen's internship:
  - the first week(s) should be used to understand the algorithm behind mapmaking
  - the hyperparameters grid exploration takes a long computational time, with a lot of waiting involved
  - in the meantime, he can understand better the subtleties of the code

• problem of IQU maps in the soft (Eva):
  - might be due to type int32 vs int64, and not shape

• CMM convolution noise (Alexandre):
  - a simple fix to the non physical ripples in the dust: call pySM at high nside, do the right convolution, and then degrade the nside
  --> it won't fix the reconstruction issue (or it shoudln't at least)
  - try to convolve a map at sigma, then epsilon, and compare it with map convolved at sigma+epsilon (for example) to see the effect of convolution with a kernel smaller than pixel size
  - vary the parameters to see which pixel size is ideal

• NMM (Leonora):
  - it will need a parameter optimisation too
  - might focus on that for now?

• problem with the fitting of r:
  - it shouldn't depend on the noise realisation
  - could it be linked to the convolution issue?
  - try with higher nside
  - try without convolution

• hysteresis in the data (Noah):
  - there was a discussion with Damien and Manuel, very interesting