[CIG-ALL] CIG WEBINAR THIS WEEK: hiPPYlib: An extensible software framework for Bayesian inversion

[Lorraine Hwang]

10 April 2015

Dear CIG Community:

Please join us this week for the continuation of our series on uncertainty quantification (UQ).  On Thursday April 14 @ 2pm PT, Noemi Petra, UC Merced will talk to us about the software tool hIPPYlib and builds on FEniCS  for the discretization of the PDE and PETSc for scalable and efficiency linear algebra operations and solvers.
hIPPYlib: An extensible software framework for large-scale Bayesian inversion

Noemi Petra, UC Merced

We present an extensible software framework for large-scale Bayesian inverse problems. This framework implements state-of-the-art scalable algorithms for PDE-based deterministic and Bayesian inverse problems and provides a unique capability for algorithmic developments for large-scale Bayesian inversion. Various features that commonly arise in these problems, such as inversions for a coefficient field and for the initial condition in a time-dependent problem are being studied in the context of an elliptic and a parabolic PDE. The derivations of the optimality conditions are shown and efficient solution algorithms are presented. The Bayesian inversion framework is stated under the  assumption of Gaussian noise and prior probability densities, and linear(ized) parameter-to-observable map. Therefore, the posterior density becomes Gaussian and can be characterized by its mean and covariance. The mean is given by the solution of a nonlinear least squares optimization problem, which we solve with an inexact matrix-free Newton-CG method. The construction of the posterior covariance is made tractable by invoking a low-rank approximation of the data misfit component of the Hessian. Short implementations of these algorithms in \Fe, which is a generic finite element  toolkit, demonstrate practical strategies for solving inverse problems with PDEs and quantifying the uncertainty associated with this inversion, reflecting the degree of "confidence'' we have in the inversion.

Connect <http://uc-d.adobeconnect.com/r28i3av93ti/> to the webinar:
	http://uc-d.adobeconnect.com/r28i3av93ti/ <http://uc-d.adobeconnect.com/r28i3av93ti/>

More details on this and future webinars can be found on our website <https://geodynamics.org/cig/events/webinars/>:
	https://geodynamics.org/cig/events/webinars/ <https://geodynamics.org/cig/events/webinars/>

Best,
-Lorraine

*****************************
Lorraine Hwang, Ph.D.
Associate Director, CIG
530.752.3656





-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.geodynamics.org/pipermail/cig-all/attachments/20160410/22b51524/attachment.html>