[SPRINT] Bayesian Foward Modeling Inference Challenge

[EiffL]

Bayesian Foward Modeling Inference Challenge

We provide a forward model from cosmology to lensing observables, your mission, if you decide to accept it, is to retrieve the best Bayesian posterior on Cosmology!

Contacts: @EiffL Time: All day from 10:30am CST Main communication channel: #desc-bayesian-pipelines-tt GitHub repo: https://github.com/LSSTDESC/bayesian-pipelines-cosmology In-person/Virtual/Hybrid: Mostly in-person and outside (in the courtyard outside of building), rain or shine. Zoom room: Usual Bayesian Pipelines Topical Team

Goals and deliverable

Goals:

• [ ] Implement a Gaussian lensing forward model
• [ ] Implement a cosmology sampler running on Perlmutter
• [ ] Ingest Gaussian Simulations from STAR challenge into the forward model

Deliverables:

• [ ] Adding Gaussian forward model to the bayesian pipeline library
• [ ] Adding a notebook demonstrating how to ingest txpipe data

Resources and skills needed

Not a lot of pre-requisites, just interest in learning more about how to implement a hierarchical forward model with a Probabilistic Programming Language.

[yesukhei-git]

This is the notebook that we worked on during the sprint day in Chicago: https://colab.research.google.com/drive/18kKAwx6ylRHg0yM1gIc7wI-sKytQwiP2?usp=sharing

1. It implements 6 cosmological parameters
2. Works only in one redshift bin
3. Produces a corner plot from the samples