In recent years, machine learning has been successfully deployed to identify phase transitions and classify phases of matter in a data-driven manner. In this talk, we present analytical expressions corresponding to the output of perfectly trained, deep neural networks at the heart of three popular methods for detecting phase transitions. The analytical expressions allow for a thorough understanding of these methods and the efficient identification of phase transitions from data without training neural networks.

Download the slides for this talk.Download ( PDF, 8178.41 MB)