Skip to content

Test different MCMC samplers and configurations for analysis #11

Description

@eduardojsbarroso

Description

When running the MCMC likelihood from Firecrown, we rely on the CosmoSIS package.
CosmoSIS offers multiple options for:

  • Ensemble samplers
  • Power spectrum generation
  • Sampler configurations and tuning parameters

At the moment, we do not know which sampler (and configuration) performs best in terms of speed and precision for our cluster likelihood.

Reference configuration:
cluster_counts_mean_mass_redshift_richness_shear.ini


Problem

We need to test different CosmoSIS samplers and configurations for cluster likelihood analyses in Firecrown. The goal is to identify the most efficient and accurate setup.

Current setup:

cosmosis cluster_counts_mean_mass_redshift_richness_shear.ini

Important: these runs should be executed using Slurm batch scripts, not directly from the terminal. Only use the test sampler on the terminal.
Checklist

  • Review available CosmoSIS samplers and their configurations
  • Write Slurm scripts to run Firecrown likelihood tests with each sampler
  • Measure runtime performance for different sampler options
  • Compare accuracy/precision of results across samplers
  • Document recommended sampler configuration for cluster likelihoods

Metadata

Metadata

Assignees

No one assigned

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions