BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//NextToU//EN
CALSCALE:GREGORIAN
BEGIN:VEVENT
UID:42bf75e6-977f-4742-b332-7302da7d3c99@nexttoyou
DTSTAMP:20261107T013850Z
DTSTART:20261117T211500Z
DTEND:20261117T221500Z
SUMMARY:Data Science Distinguished Lecture: Agostino Capponi - Industrial 
 Engineering and Operations Research\, Columbia University
LOCATION:Frank H. T. Rhodes Hall 253
DESCRIPTION:This talk is co-sponsored with the ORIE Colloquium and the Cen
 ter for Data Science for Enterprise & Society.\n\nTALK: Data-Driven Learn
 ing of Latent Dynamics\n\nAbstract: High-dimensional time series often ex
 hibit complex dependencies that can be explained by a few latent factors.
  Uncovering these factors and modeling their dynamics is a fundamental ch
 allenge in data science\, with applications ranging from signal processin
 g and forecasting to asset pricing and financial risk management.\n\nWe i
 ntroduce a data-driven framework for modeling and predicting a multivaria
 te response y(t) from a high-dimensional covariate process x(t)\, allowin
 g for nonlinear dependencies without imposing a parametric model on the c
 ovariate dynamics. We leverage diffusion maps to uncover low-dimensional 
 representations of the covariate process that retain explanatory power fo
 r the response time series (y(t)). We establish concentration inequalitie
 s that provide convergence guarantees for time averages and show that\, u
 nder mild assumptions\, the resulting diffusion map embeddings exhibit ap
 proximately linear dynamics. This property enables us to develop a geomet
 ric Kalman filter that operates in the learned embedding space to estimat
 e latent states and predict the response variable.\n\nWe demonstrate the 
 effectiveness of our framework in financial stress testing. Using histori
 cal data spanning three major financial crises\, we show that our data-dr
 iven approach reduces the mean absolute error of scenario-based \n\nhttp:
 //next-to-u.app/events/42bf75e6-977f-4742-b332-7302da7d3c99
URL:http://next-to-u.app/events/42bf75e6-977f-4742-b332-7302da7d3c99
END:VEVENT
END:VCALENDAR