Want a framework for flexible joint Bayesian modeling of multiple time series where inference is fast, easy, and scalable?

The second June seminar of the Center for Quantitative Methods and Data Science (QM&DS), in partnership with the Biostatistics, Epidemiology and Research Design (BERD) Center at Tufts CTSI and the Data-Intensive Studies Center (DISC) at Tufts University, is Wednesday, June 16, 2:00-3:00PM via Zoom. The topic is Scalable Bayesian Flexible Joint Time Series Modeling, presented by Michael Wojnowicz, PhD.

We frequently obtain datasets containing multiple time series — that is, a collection of sequences, often corresponding to temporal data from multiple individuals. For example, consider movement patterns of soldiers during a ruck march, lesion counts of multiple sclerosis patients, or computer activity by employees at a company. In this talk, we describe a framework for flexible joint Bayesian modeling of multiple time series where inference is fast, easy, and scalable. In particular, we construct a scalable Bayesian approach to mixed HMMs, where mixed HMMs are Hidden Markov Models with multi-level generalized linear models (a.k.a. Generalized Linear Mixed Models, or Mixed Effects Models) embedded within the transitions and emissions structure. Mixed HMMs are an excellent framework for personalized time series modeling: models can be personalized, while sharing statistical strength across individuals to “fill in” knowledge as necessary, based on knowledge about other individuals, and particularly similar individuals. Moreover, the impact of dynamic covariates can be learned based on their effects across the entire population of individuals. In this talk, we will introduce mixed HMMs, and then discuss how to make inference fast, easy, and scalable.


Michael Wojnowicz, PhD is a Data Scientist II in the Data Intensive Studies Center (DISC) at Tufts University, working with the Machine Learning Research Group. He earned his Ph.D. from Cornell University in 2012, where his work in Cognitive Science led to the Dallenbach Fellowship for Research Excellence, the Cognitive Science Dissertation Proposal Award, and the Cognitive Science Graduate Research Award. Dr. Wojnowicz also has master’s degrees in Mathematics (University of Washington) and Statistics (University of California at Irvine). Before joining Tufts University, Dr. Wojnowicz was the Distinguished Data Scientist at Cylance. At Cylance, he developed statistical machine learning models for detecting malicious computer files and anomalous user activity, leading to 10 patents (5 granted, 5 pending). Dr. Wojnowicz’ current research interests include time series modeling, variational inference, and nonparametric Bayesian modeling.


Wednesday, June 16, 2:00-3:00PM, via Zoom


To attend, please register here via Tufts CTSI I LEARN.