Novel reduced order modeling strategies for solving stochastic inverse problems

We develop and implement data-driven, projection-based methods for reducing computational demands during the solution of stochastic inverse problems.

Coarse mesh FE heat eqn. soln.

Ground truth soln.

https://earls.cee.cornell.edu/wp-admin/post.php?post=101&action=edit#

POD enriched GFEM coarse mesh solution

Manifold interpolation to uncover reduced order sub-spaces