PhD Candidate
Georgia Institute of Technology
Areas of Interest
- Artificial Intelligence
- Information, Data, Network, and Communication Sciences
- Signal Processing
- Sequential Change-point Detection
- Robust Hypothesis Testing
- Distributionally Robust Optimization
Poster
Robust Statistical Inference Through the Lens of Optimization
Abstract
Modeling, learning, and making inferences from high-dimensional streaming data is an important topic broadly interested in statistics, computer science, and machine learning. Modern data acquisition calls for new theoretical foundations to guide the development of online algorithms. Our work is motivated by two major challenges in sequential statistical inference. First, we may have limited memory due to storage ability or privacy constraints. For the dynamic situation where the underlying distribution of data samples changes over time, we need to consider the trade-off between the memory and the minimal information we need. Second, we may have very limited data samples from one particular class, which results in non-negligible uncertainties for the estimation procedure that may affect the reliability of decisions we make.
This presentation will highlight our recent efforts to make robust statistical inference by combining techniques from both statistics and optimization. In particular, the first part of the presentation will focus on two theoretical developments: one is the sequential change detection for multi-sensors, assuming the change happens to the covariance structure of the sensor network; the other is robust hypothesis testing that hedge against the distributional uncertainties in data-driven scenario. The second part of this presentation will provide a variety of applications of sequential change detection, such as seismic tremor detection human activity detection, as well as new opportunities and challenges in this field.
Bio
Liyan Xie is a fifth-year Ph.D. Candidate at Georgia Institute of Technology in the H. Milton Stewart School of Industrial and Systems Engineering. She received her B.Sc. in Statistics from University of Science and Technology of China (USTC) in 2016. Her research interests are the intersection between statistics and optimization, with a primary focus on sequential change detection and their applications in sensor networks and industrial quality control. She has worked on sequential detection, robust hypothesis testing, and spatio-temporal modeling in seismic data analysis and wearable sensors. She was awarded the finalist of INFORMS QSR (Quality, Statistics, and Reliability) student paper competition and Runner-up for the INFORMS Computing Society Student Paper Prize in 2019. She received the 2020 ARC student fellowship and IDEaS-TRIAD Research Scholarships from the Transdisciplinary Research Institute for Advancing Data Science (TRIAD) at Georgia Tech.
