-
Notifications
You must be signed in to change notification settings - Fork 0
Home
Stéphane Derrode edited this page May 5, 2026
·
3 revisions
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.
| 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 |
-
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).
See Citing or the CITATION.cff at the repo root.