Speaker
Jack Lidmar
(KTH Royal Institute of Technology)
Description
In computer simulations, using molecular dynamics or Monte
Carlo, it is often the case that it is harder to converge
the simulation for certain parameters than other. We
discuss how the notion of distance in a parameter space of
probability distributions can help deciding and optimizing
how samples should be distributed in the parameter space.
In particular, we show how a properly defined metric may be
used to guide extended ensemble simulations over bottleneck
configurations in the parameter space, and how this may
improve, e.g., biomolecular simulations.