experiments list (for the least publishable unit)

[humnaawan]

The goal is here to create a list of experiments to be run for the least publishable unit for the CMB cosmology inference using SBI and MCMC. Here are some initial ideas:

• Impact of which parameters we constrain
  For example, we can fix one of r, A_s or constrain both. Functionality to vary only these two cosmology parameters is currently available via DeepCMBsim.

• Impacts of the noise model
  DeepCMBsim has poisson noise but we could construct toy models for more complicated models, e.g., those that couple different modes.

• Galactic foregrounds
  These are most intuitive to deal at the map level -- which may require development in DeepCMBsim -- but we could proceed with e.g., complex noise models that capture the essence of the foregrounds.
  We can think about whether we still want to do a 2-point analysis (i.e., the data vector will be a power spectrum; this will be the simplest way to do a 1-1 comparison between SBI and MCMC) or if we want to work with maps right away (on top of a 2-point analysis).

Let's use this thread to discuss things and finalize the experiments list for the first least publishable unit, alongside keeping track of any ideas for future R&D.

[bnord]

note: $A_s$ is not a cosmology parameter, it's a kind of noise or nuisance parameter.

[bnord]

To design our experiments, we'll need to have a clear picture of the landscape of different modeling methods, including the quadratic estimator.

Related key issue: #4