triplet-fitting: Gaussian mixture models of nonlinear distortion

Python · MATLAB engine · scikit-learn · pandas · research code · GitHub · DOI 10.5281/zenodo.13628611

In a nonlinear fibre, the distortion of a symbol depends on its neighbours, so the received points are not the simple Gaussian clouds that standard receivers assume. This package builds datasets of symbol triplets from simulated transmissions (generated with hpcom) and fits mixtures of Gaussians to them.

• Dataset builders for triplets across launch powers and independent runs, and a fit_all entry point to fit them in bulk.
• EM fitting of Gaussian mixtures through the MATLAB engine and David Barber’s BRML toolkit, wrapped behind a Python interface.
• Packaged, tested and archived on Zenodo with a citable version.

Related paper: Examining nonlinear behaviors in optical communication systems using Gaussian mixture model.