Physics Colloquium: “Unsupervised Machine Learning of Quantum Phase Transitions”

Zhexuan Gong

Colorado School of Mines, Physics Department

Abstract:  Experimental quantum simulators have become large and complex enough that discovering new physics from the huge amount of measurement data can be quite challenging, especially when little theoretical understanding of the simulated model is available. Unsupervised machine learning methods are particularly promising in overcoming this challenge. I will review typical unsupervised learning methods and show that they generally only work for learning simple symmetry-breaking quantum phase transitions. I will then show that a more advanced method known as diffusion map, which performs nonlinear dimensionality reduction and spectral clustering of the measurement data, has much better potential for unsupervised learning of complex phase transitions, such as topological phase transitions and many-body localization. This method is readily applicable to many experimental quantum simulators as it only requires measuring each particle in a single and local basis.

Reference: A. Lidiak and Z.-X. Gong, Phys. Rev. Lett. 125, 225701 (2020).

Biography: Zhexuan Gong obtained his PhD in physics from University of Michigan in 2013. He then spent four years working as a research associate and a research scientist at the Joint Quantum Institute associated with the University of Maryland and the NIST. He joined Mines as an assistant professor of physics in 2018 and also holds an associate position at NIST. Gong’s research is in the field of theoretical quantum physics. In particular, he is interested in designing faster architectures for quantum computing, understanding novel quantum many-body physics via quantum simulation, and applying machine learning to facilitate new discoveries in quantum experiments. He recently won a prestigious W. M. Keck Foundation award to support his research in building a next-generation quantum simulation platform. In his free time, he enjoys playing piano at home, especially with his 20-month old daughter sitting on his lap.

Unless specifically noted, the talks will all be held in CoorsTek 140/150