16.940 Numerical Methods for Stochastic Modeling and Inference
Fall 2018
Instructor: Youssef M Marzouk
Lecture: TR1-2.30 (35-225 [note updated room])
Information:
This subject is an advanced introduction to numerical methods for treating uncertainty in modeling and simulation. Examples will be motivated by engineering and science applications, but the techniques we study will be more broadly applicable. In a sense, this is a subject about the field of uncertainty quantification, with a heavy dose of computational statistics.
Part one focuses on uncertainty propagation and assessment, with foundations in Monte Carlo simulation and in approximation theory: Monte Carlo methods; variance reduction; global sensitivity analysis; polynomial approximation; Gaussian process regression and radial basis function approximation; stochastic Galerkin and collocation methods; sparse grids and other methods for high-dimensional approximation.
Part two focuses on the interaction of models with observational data, from a statistical perspective: Bayesian parameter inference; inverse problems; prior modeling; Markov chain Monte Carlo; sequential Monte Carlo; nonlinear filtering and data assimilation; model selection; model error and model validation.
Announcements
No announcements
