This event has passed.
Thursday, January 23, 2025
2:00 pm – 3:00 pm
Presenter: Joe Qin is the Wai Kee Kau Chair Professor of Data and President of Lingnan University in Hong Kong.
Abstract:
Multi-dimensional time series are ubiquitous in engineering, science, and economics. While the dimension of sensors increases with modern sensing technology and data acquisition, the dimension of dynamics is often relatively small. In this talk I will present a probabilistic latent vector autoregressive framework, known as Principal Predictor Analysis (PPA), where dimension reduction and optimal dynamic prediction are simultaneously achieved. The PPA solution is a dynamic parallel to principal component analysis. The dynamic latent variables are related to a reduced dimensional predictor with maximized predictability. An iterative solution is developed using a maximum likelihood framework. Data from a Lorenz oscillator and an industrial process are used to show the superiority of the proposed algorithm. The reduced-dimensional dynamic modeling framework has potentially wide applications in prediction, control, and diagnosis of anomalies.
