analysis workflow

[humnaawan]

Our overarching goal of this repo to host an inference pipeline for CMB cosmology - via two inference methods, specifically. The following workflows capture the v0 plan (developed during the DeepSkies retreat 04/2023)

We'll use this thread to discuss iterations on these workflows.

Note: the PS_Maker is the power spectrum object from DeepCMBsim.

[humnaawan]

Notes from 9/27 meeting:

• differences between the two workflows? data and theory noise models in SBI workflow should be connected.
• workflows would look different for maps vs cls -- to be discussed further when discussing the experiments list
• raw data or spectrum: working with maps could make systematics handling and mitigation easier but it makes 1-1 comparison with MCMC difficult (unless e.g. we do a beyond 2-pt analysis with MCMC).

@bnord @cavestruz please add/edit anything I might have missed.

[humnaawan]

here's the updated SBI workflow, now analogous to the MCMC workflow in its direct connection between theory and data (via noise):