NSIW2023-keynote

Companion repository for the keynote talk titled "Variational Bayesian inference for System Identification" at the Nonlinear System Identification Workshop 2023 in Eindhoven.

Demonstration

The code folder contains a Jupyter notebook detailing the demonstration system along with julia code and figures. To run the notebook, install Jupyter, download a julia kernel (v1.8+), and run:

using Pkg
Pkg.add("IJulia")

That will add the julia kernel to jupyter automatically. Alternatively, consider installing Visual Studio Code and adding the Jupyter and Julia extensions.

If you're not interested in running the notebook, but would just like to view it, open the .html version in your browser. It should render with all the figures and results.

References:

• Kouw, Podusenko, Koudahl & Schoukens (2022). Variational message passing for online polynomial NARMAX identification. American Control Conference, DOI: 10.23919/ACC53348.2022.9867898.
• Podusenko, Akbayrak, Senöz, Schoukens & Kouw (2022). Message passing-based system identification for NARMAX models. IEEE Conference on Decision & Control, DOI: 10.1109/CDC51059.2022.9992891.
• Senöz, van de Laar, Bagaev & de Vries (2021). Variational message passing and local constraint manipulation in factor graphs. Entropy, DOI: 10.3390/e23070807.
• Van de Laar (2021). Chance-constrained active inference. Neural Computation, DOI: 10.1162/neco_a_01427.
• Dauwels (2007). On variational message passing on factor graphs. IEEE International Symposium on Information Theory, DOI: 10.1109/ISIT.2007.4557602.