MEM Seminar Thursday 5/22 10:30am: Pierre Lermusiaux MIT

[M. Ani Hsieh]

All,

This is a talk that may be of interest.  Let me know if you want to meet with the speaker.

Thanks.

Ani

---------- Forwarded message ----------
From: M. Ani Hsieh <mhs...@drexel.edu>
Date: Mon, May 19, 2014 at 4:54 PM

MEM Seminar Series

Thursday May 22, 2014

Location/Time: Curtis 162 (MEM Seminar Room) – 10:30 am

 

“Bayesian Inference of High-Dimensional Dynamical

Model Formulation”

Dr. Pierre F.J. Lermusiaux

Associate Professor of Mechanical Engineering

Massachusetts Institute of Technology

 

Abstract: In this presentation, we address a holistic set of challenges in high-dimension ocean Bayesian nonlinear estimation: i) predict the probability distribution functions (pdfs) of large nonlinear ocean systems using stochastic partial differential equations, ii) assimilate data using Bayes' law with these pdfs, iii) predict the future data that optimally reduce uncertainties, and (iv) rank the known and learn the new model formulations themselves. Overall, we allow the joint inference of the state, equations, geometry, boundary conditions and initial conditions of dynamical models. Examples are provided using time-dependent fluid and ocean flows, including cavity, double-gyre and sudden-expansion/Strait flows with jets and eddies. The Bayesian model inference, based on very limited observations, is illustrated by the estimation of obstacle shapes and positions, and of biogeochemical reaction equations. These estimation problems have interesting links to classic mechanical engineering. A third estimation, the inference of multiscale bottom gravity current dynamics, is motivated in part by classic over

--
--
M. Ani Hsieh, Ph.D.
Assistant Professor
Mechanical Engineering & Mechanics
Drexel University
Email: mhs...@drexel.edu
Web: http://drexelsaslab.appspot.com/
Phone: 215-895-5870/5803