Physical Interpretation of a Generated Class-Label Vector by GK-Means in Spectral Width and Mean Power Analysis of Supercontinuum

The polarization state of an erbium-doped fiber laser was changed, and its dependence on the spectral width and mean power generated at the end of a high-nonlinear fiber was investigated. The GK-Means algorithm was applied to identify clustering patterns in the spectral width and mean power. Finally, linear polarized vectorial solitons were simulated as the input of the high-nonlinear fiber, providing a physical interpretation of the clusters proposed by GK-Means. A comparison of the simulation and experimental data revealed approximately 42% of coincidences.