iter-inla

A framework for state and parameter estimation in nonlinear dynamical systems using iterated INLA.

Overview

This repository contains the accompanying implementation for the UAI paper "Iterated INLA for State and Parameter Estimation in Nonlinear Dynamical Systems".

Dependencies

• NumPy
• SciPy
• Matplotlib
• scikit-learn
• scikit-sparse fork with Takahashi equations
• findiff fork with periodic boundary conditions

Installation

This Python module depends on the suite-sparse library, which can be installed with your package manager of choice. For example, with conda we can install suite-sparse via:

# Install suite-sparse with conda
conda install -c conda-forge suitesparse

We also need custom forks of the scikit-sparse and findiff packages, which can be installed via:

# Install scikit-sparse fork
git clone https://github.com/rafaelanderka/scikit-sparse.git
pip install ./scikit-sparse

and

# Install findiff fork
git clone https://github.com/rafaelanderka/findiff.git
pip install ./findiff

Finally, we can install iter-inla with

# Install iter-inla
git clone https://github.com/rafaelanderka/iter-inla.git
pip install ./iter-inla

The module can then be imported in Python as iinla.

Usage

To get started, please have a look at the demos directory, which provides examples with live preview plots.

Citing

If you found this useful, please consider citing:

@article{anderka2024iterated,
  title={Iterated {INLA} for State and Parameter Estimation in Nonlinear Dynamical Systems},
  author={Anderka, Rafael and Deisenroth, Marc Peter and Takao, So},
  journal={Proceedings of the Fortieth Conference on Uncertainty in Artificial Intelligence},
  year={2024},
  publisher={PMLR}
}

License

MIT