Speaker
Description
We present an open source Python package to compare simulation-based inference (SBI) approaches to MCMC inference in the context of galaxy cluster mass estimates from gravitational weak lensing data. The package, CLSBIWeakLens, provides a modular framework to flexibly run numerical experiments on cluster mass estimation from radial profiles, a typical data vector in optical cluster cosmology from surveys such as the Vera C. Rubin Observatory Legacy Survey of Space and Time (LSST). CLSBIWeakLens has modules to define the ‘simulations’ used to train the SBI, modules to build the inference procedures, and a straightforward interface for the user to create new experiments through user-defined configuration files. Our framework allows for testing the robustness of inference posteriors to noise in the data, approaches to stack data vectors to average out the noise, and modeling choices for population statistics and the mass distribution within each galaxy cluster. We illustrate example experiments that we lay out in tutorials, highlighting (1) tests on the effects of noisy data on posteriors with inference performed on stacked radial profiles and on individual radial profiles, and (2) tests on inference performed with a misspecified model. This framework provides foundations to the potential application of SBI to speed up inference in hierarchical models for galaxy cluster cosmology. The source code is publicly available on GitHub.
| Would you be interested in presenting a poster if the conference is oversubcribed? | Yes |
|---|
