September 10, 2021

About the speaker

Dr. Xianqi Li is an assistant professor at Florida Institute of Technology. He was a Postdoctoral Research Fellow in the Department of Radiology at Massachusetts General Hospital within Harvard University. He received his Ph.D in Mathematics at the University of Florida in 2018 and his MS in Mathematics from the University of Texas Pan-American in 2009. Research Interests include Optimization, Image Processing, Machine Learning, Deep Learning, and Data Analytics.

Data-Driven Modeling, Optimization and Prediction: Critical Factors for Decision Making

As the amount of available data grows exponentially, the decision-making process for groups, organizations and even individuals is being changed. For instance, companies can decide which types of products to create or increase to best meet customers' needs by gathering data online through their buying habits instead of relying on the lengthy process of customer feedback. In healthcare, doctors are able to make better treatment decisions for patients by collecting enough personal data such as lifestyle, disease history and genetics on each person. However, it is not the amount of data that only matters. The approaches we do the data modeling, solve the formulated optimization problem and make predictions are all critical components for decision making. In this talk, I will show how these components driven by big data aid us make decisions. First, I will present causality network learning with temporally dependent data by introducing our proposed rigorous and computationally efficient statistical machine learning methods. I will talk briefly about the established asymptotic upper bound on the estimation error rates of the introduced models and the convergence rate of the designed optimization algorithms, which will reveal to us how the data size influences the performance of the modeling and algorithm. Then, I will discuss the partially parallel MR image reconstruction problem from optimization perspective and its significance in healthcare. Lastly, I will present our recently developed deep learning models and show how the doctors benefit from them by combining with the large amount of healthcare data.