Tensor Learning Team Seminar (Talk by Dr. Shi-Ju Ran, Physical Department, Capital Normal University, China).

Speaker: Dr. Shi-Ju Ran, Physical Department, Capital Normal University, China

Title: Tensor Network for Efficient and Interpretable Machine Learning

Date: July 7, 2023

Time: 3:00 pm - 4:00 pm (JST)

Venue: Online

Abstract:

In this talk, I will introduce our recent progresses in the machine learning models and algorithms based on tensor network (TN) from the following two aspects. First, TN can be regarded as a generalization of neural network (NN). A residual net formed by tensor blocks, named as residual matrix product state, was proposed for supervised learning. Meanwhile, we show that TN can be used as an efficient representation of high-order tensors that “exponentially” lowers the complexity. This conceived a powerful scheme for NN compression. Second, we show the interpretability of TN machine learning based on the quantum probabilistic interpretation. The theories of quantum entanglement and measurement were used to develop machine learning schemes for compressed sampling and feature selection.

Bio: Shi-Ju Ran is currently an associated professor in the Department of Physics, Capital Normal University, China. He received the Ph.D. degree in University of Chinese Academy of Sciences in 2015, and then joined ICFO - the Institute of Photonic Sciences, Spain, as a post-doctoral researcher for the next three years. His research interests include tensor network methods, quantum machine learning, quantum computation, and quantum many-body physics.