A Julia package featuring various ABC methods.
This package implements
-
Classic ABC: See the historical paper Pritchard et al. - 1999
-
ABC-SMC: A sequential Monte Carlo version of the classic ABC method Moral et al. - 2012. Note that this is the only method not using a reference table for training.
-
ABC-CNN: ABC method as described by Åkesson et al. - 2021.
-
ABC-Conformal: ABC method completely free of summary statistics and threshold selection as described in Baragatti et al. - 2024. It use Approximate Bayesian Computation (ABC) with deep learning and conformal prediction.
The documentation is not (yet) available however, detailed Quarto notebook on several examples using all implemented ABC methods are available.
- MA(2) example (2 parameters) -> Well known toy Bayesian example.
- Discrete Lotka-Volterra example (3 parameters) -> Very challenging for some extreme parameters.
- Phytoplankton dynamics in Lake with a toy model example (9 parameters) -> High dimensional example.
The package is currently NOT in the Julia general registry (will be soon), but on a local registry, hence to download it just add:
using Pkg
pkg"registry add https://github.com/dmetivie/LocalRegistry"and then add it as a normal package
Pkg.add("ABCMethods")
# or
pkg> add ABCMethods