Welcome to desilike’s Documentation!

desilike is a modern cosmological inference framework used by the Dark Energy Spectroscopic Instrument (DESI) experiment. While initially designed for DESI, it is a stand-alone inference code that can also be used with other cosmological datasets. Conversely, desilike also supports exporting DESI likelihoods for use in other inference codes such as Cobaya, CosmoSIS, or MontePython.

Key features:

• Advanced parameterization infrastructure (priors, reference distributions, derived parameters, etc.)

• MPI support for parallel posterior computations

• Consistent parameterization between Boltzmann codes (via cosmoprimo)

• Transparent namespace scheme for parameters

• In-place emulation for faster inference

• General differentiation routines (with JAX for automatic differentiation)

Table of Contents

User Documentation

• Building
• Getting started

API Reference

• Core
• Theories
• Observables
• Likelihoods
• Bindings
• Profilers
• Samplers
• Samples
• Emulators

Development

• Introduction
• Documentation
• Unit Tests