Snake net with a neural network for detecting multiple phases in the phase diagram

Unsupervised machine learning applied to the study of phase transitions is an ongoing and interesting research direction. The active contour model, also called snake was initially proposed for target extraction in two-dimensional images. In order obtain a physical diagram, model with artificial neural network unsupervised way by authors [Phys.Rev.Lett. 120, 176401(2018)]. It guesses boundary as initial then drives convergence forces estimated network. this paper, we extend method one net multiple contours purpose obtaining several boundaries diagram. For classical Blume-Capel diagram containing three four phases obtained. Moreover, overcome limitations position speed up movement snake, balloon force decaying iteration steps introduced structure. Our helpful determining phases, using just snapshots configurations from cold atoms or other experiments without knowledge phases.