Bayesian Variational Inference for Nonlinear Dynamical Systems in Julia
This package implements an extended Variational Laplace routine for parameter inference in nonlinear models, including support for:
- Thermodynamic integration for accurate model evidence estimation
- Low-rank covariance structure for scalable inference
- Adaptive noise modeling (heteroscedastic variance)
- Flexible support for arbitrary model functions
Originally developed for use in computational neuroscience, this general-purpose inference engine can be used in any context where you want to infer latent parameters from noisy observations generated by a nonlinear forward model.
using Pkg
Pkg.add(url="https://github.com/alexandershaw4/VariationalLaplace.jl")
Alexander Shaw 2025. https:cpnslab.com