Feature Tracker

[benjaminpope]

Here are some features we need to implement:

Basic Features

• Visibility functions for rings, gaussians (More Source models #22)
• Improve syntax for vmapped functions
• Model evidences by Laplace
• Monte Carlo injection-recovery
• ROC Curves from evidence / injection recovery
• Automatic plotting functions
• Interface to MCMC libraries (eg BlackJax, NumPyro) (MCMC and VI implementations #12)

API Features

• Unify APIs of all likelihood functions (eg what is a contrast?)
• Model specification by equinox object
• Hierarchical model specification (Issue Hierarchical Systems of Model Components #20)
• Interface with dLux (More Source models #22)
• Interface with Harmonix (More Source models #22)
• interface with @shashankdholakia's ELR models (More Source models #22)
• Fix spaghetti code calling implaneia for IO, simplify dependencies
• Rationalize functions and classes appropriately into models.py, inference.py, utils.py etc
• Fix convention about jnp, np vs np, jnp replace onp, np with np, jnp #21

Unit Tests

• Unit tests for gridding
• Unit tests for plotting

Docs

• Document all visibility functions
• Document all likelihood functions
• Document data IO
• Document plotting functions

Examples

• Comparison of contrast detection limits with Absil, Ruffio, UQ methods
• Contrast curve for non-binary source
• MCMC fitting example (MCMC and VI implementations #12)
• Fix data directory to have something attractive + public there for examples and unit tests

[benjaminpope]

vmap syntax, object-oriented model specification, automatic plotting code, unit tests, and unification of likelihoods to use flux (not contrast) completed in PR #11.