Skip to content
Stéphane Derrode edited this page May 5, 2026 · 3 revisions

exactIMM — Wiki

Welcome to the exactIMM wiki. This software accompanies the paper

On Fast Optimal Filtering in Gaussian Switching Systems
S. Derrode & W. Pieczynski (2026)

and provides a reference implementation of the H5-constrained Gaussian Switching System framework.

Where to start

If you want to… go to
Install and run a first example Installation
Walk through a complete simulate→filter→learn cycle Tutorial
Reproduce the paper figures and tables Paper-Reproduce
Understand the codebase architecture API-Overview
Use the GUI to explore parameters interactively GUI-Guide

What is in the package?

  • Filter (prg/filter/) — exact constant-gain optimal filter under (H5), plus a general IMM mode for unconstrained models.
  • Simulator (prg/simulate.py) — fast iterator-based generator of GSS trajectories.
  • Learning (prg/learning/) — supervised OLS and semi-supervised Baum-Welch EM with three H5 projection variants ((\tau \in {B, A^\dagger, \Sigma_U})).
  • Experiments (prg/experiments/) — Monte-Carlo benchmarks (§6 of the paper) and the ENSO real-data study (§7).
  • GUI (prg/gui/) — interactive PyQt6 interface (optional).

Citing

See Citing or the CITATION.cff at the repo root.

License

GNU Affero General Public License v3.0

Clone this wiki locally