Title:
Similarity metrics for physical simulation verification and optimization problems
<br><br>
Abstract:
Quantification of similarity is a challenging problem for problems
involving high dimensional data, especially when the data has a
context (such as a physical system) on which the notion of similarity
is highly dependent.  This talk will cover topics related to the
construction, application, and understanding of similarity metrics in
the context of computer simulations in the physical sciences.  These
metrics are used for both validation of simulations against
experimentally obtained data, and as distance functions for use in
optimization problems in which the simulation plays the role of
objective function.  The models that motivate the work that will be
presented in the talk produce output in the form of image or
volumetric data, and the properties of the data that are captured by
useful metrics are geometric in nature.