Mar 2 (Thu) @ 11:00am: "Research on Tensor Optimization and Applications in Sustainable AI," Zi Yang, Postdoc, UCSB

Zoom Meeting – https://ucsb.zoom.us/j/86771502945

Abstract

As a higher-order generalization of matrices, tensors are natural to represent and process multi-dimensional data arrays and hence are widely applied to quantum physics, scientific computation, uncertainty quantification, machine learning, and many other fields.

In the talk, I will discuss my research on the theories of tensor computation and applications of tensors in sustainable AI. The first part of the talk will be about the research on fundamental theories of tensor properties. Detecting copositivity for symmetric tensors and properties of Hermitian tensors will be covered in the first part. In the second part of the talk, I will introduce the tensor applications in sustainable AI, including a mixed-precision stochastic algorithm for CP tensor decompositions and how to compress large-scale neural networks for training and deployment on resource-constrained devices via tensor decompositions.

I will conclude by outlining future research directions in bridging tensor theories and machine learning, tensor learning algorithm design for sustainable AI, and theoretical analysis of tensor-compressed AI models.

Bio

Zi Yang is a postdoctoral scholar in the Department of Electrical & Computer Engineering at the University of California, Santa Barbara. His research focuses on 1) fundamental theories of tensor properties and tensor computation; 2) analysis and design of sustainable tensor learning algorithms for AI. Before joining UC Santa Barbara, he received his Ph.D. in mathematics from UC San Diego, with a focus on tensor computation and optimization.

Hosted by: Professor Zheng Zhang

Submitted by: Zi Yang <ziy@ucsb.edu>