DMS Graduate Student Seminar

Time: Feb 22, 2023 (03:00 PM)
Location: 108 ACLC

Speaker: Dr. Rob Molinari 
Title:  Robust and Scalable Inference for Stochastic Processes
Abstract: Stochastic processes are broadly used in many applications and fields of research going
from astronomy, engineering, and physics to healthcare and social sciences. One of the main purposes for which they are used is to obtain inference on parameters, interpretable predictions, and reliable uncertainty quantification. Their use is however challenged by the growing size of data and the presence of different forms of contamination in parts of the data (including missingness) which severely limit the use of commonly employed statistical approaches. To address this, we make use of a moment-based method built upon a wavelet-decomposition of the data where it is possible to adapt and limit the impact of these different forms of contamination while preserving reasonable computational efficiency as data scales. We present the theoretical and applied properties of this framework and discuss the ongoing extensions to different tasks of robust inference on large-scale stochastic processes.