Samplers

Samplers are the different methods that CosmoSIS uses to choose points in parameter spaces to evaluate.

Some are designed to actually explore likelihood spaces; others are useful for testing and understanding likelihoods.

Simple Samplers

• test: Evaluate a single parameter set and save all results
• list: Evaluate a pre-made list of parameter sets

MCMC Samplers

• emcee: Ensemble walker sampling
• metropolis: Classic Metropolis-Hastings sampling
• importance: Importance sampling
• kombine: Clustered KDE
• pmc: Adaptive Importance Sampling
• pocoMC: Preconditioned Monte Carlo for Accelerated Posterior and Evidence Estimation

Nested Samplers

• multinest: Nested sampling for Bayesian Evidence
• polychord: Ensemble nested sampling for Bayesian Evidence
• nautilus: ML-enhanced nested sampling for Bayesian Evidence

Optimizers

• maxlike: Find the maximum likelihood using various methods in scipy
• minuit sampler MPI-aware maxlike sampler from the ROOT package.
• gridmax: Naive grid maximum-posterior

Grid Samplers

• grid: Simple grid sampler
• snake: Intelligent Grid exploration

Specialist Samplers

• fisher: Fisher matrix calculation
• star: Simple star sampler
• apriori: Draw samples from the prior and evaluate the likelihood