Tensor Learning Team Seminar (Talk by Dr. Pan Zhang, Chinese Academy of Sciences).

Title: Tensor networks for unsupervised machine learning

Speaker: Dr. Pan Zhang, Institute of Theoretical Physics, Chinese Academy of Sciences (ITP, CAS)

Abstract:
Modeling the joint distribution of high-dimensional data is a central task in unsupervised machine learning. In recent years, many interests have been attracted to developing learning models based on tensor networks, which have advantages of theoretical understandings of the expressive power using entanglement properties, and as a bridge connecting the classical computation and the quantum computation. Despite the great potential, however, existing tensor-network-based unsupervised models only work as a proof of principle, as there is a significant gap in the performance between the tensor-network models the standard models such as the restricted Boltzmann machines (RBM) and neural networks.

In this talk Pan Zhang will first present how to understand the gap using an exact two-dimensional representation of RBM, then present a new tensor-network-based model, the Autoregressive Matrix Product States (AMPS), which combines the matrix product states from quantum many-body physics and the autoregressive models from machine learning. The model enjoys exact calculation of normalized probability and unbiased sampling, as well as a clear theoretical understanding of expressive power. We demonstrate the performance of the AMPS using two applications, the generative modeling on synthetic and real-world data, and the reinforcement learning in statistical physics. Using extensive numerical experiments, we show that the proposed model significantly outperforms the existing tensor-network-based models and the restricted Boltzmann machines, and is competitive with the state-of-the-art neural network models.

Bio: Pan Zhang is a professor at the Institute of Theoretical Physics, Chinese Academy of Sciences (ITP, CAS). He finished his Ph.D. at Lanzhou University and ITP, CAS in 2009 and did post-docs at spin glass theory groups in Europe and the Santa Fe Institute in the USA before joining ITP, CAS in 2015. Pan Zhang’s research interest is in the interdisciplinary field of statistical physics, machine learning, quantum many-body, and quantum computation.