CDS 270-3

References
Lectures

Course Description

CDS 270-3. Stochastic System Analysis and Bayesian Updating. This course is taught concurrently as AM 125c. Units 3-0-6.

Course Outline

This course focuses on a probabilistic treatment of uncertainty in modeling a dynamical system's input-output behavior, including propagating uncertainty in the input through to the output. It covers the foundations of probability as a multi-valued logic for plausible reasoning with incomplete information that extends Boolean logic, giving a rigorous meaning for the probability of a model for a system. Approximate analytical methods and efficient stochastic simulation methods for robust system analysis and Bayesian system identification are covered. Topics include: Bayesian updating of system models based on system time-history data, including Markov Chain Monte Carlo techniques; Bayesian model class selection with a recent information-theoretic interpretation that shows why it automatically gives a quantitative Ockham's razor; stochastic simulation methods for the output of stochastic dynamical systems subject to stochastic inputs, including Subset Simulation for calculating small "failure" probabilities; and Bayes filters for sequential estimation of system states and model parameters, that generalize the Kalman filter to nonlinear dynamical systems.

Tuesday, Thursday 9 AM - 10:30 AM; 306 Thomas

James L. Beck
Office hours: by appointment

David Pekarek
Office hours: by appointment

Announcements

• The first lecture will be on Tuesday, March 30, 2010 in 306 Thomas at 9 AM.

Course Material

• References