Machine learning classification of two-dimensional vortex configurations

We consider computer generated configurations of quantised vortices in planar superfluid Bose-Einstein condensates. We show that unsupervised machine learning technology can successfully be used for classifying such vortex configurations to identify prominent vortex phases of matter. The machine learning approach could thus be applied for automatically classifying large data sets of vortex configurations obtainable by experiments on two-dimensional quantum turbulence.