| Abstract |
|
Dynamic (also known as Benamou-Brenier) formulations of optimal
transport (OT) and its extension to the unbalanced setting, frame the
task of comparing two given positive measures as a variational problem
in which one searches for measure paths that minimize a certain
kinetic energy functional. In applications, it is frequently natural
to require such paths to satisfy additional conditions. Inspired by
this, we introduce a model which incorporates, into the framework of
unbalanced OT, a particular family of integral equality and inequality
constraints on the admissible paths of the distribution, transport
field and/or source term. We examine the well-posedness of the
resulting class of convex variational problems and the properties of
the induced metrics on the corresponding measure subspaces. We will
then consider a discretized version of those problems and derive an
adaptation of the parallel proximal scheme from convex optimization to
numerically approximate solutions.
|
For future talks or to be added to the mailing list: www.math.uh.edu/colloquium