Bi-Co Math Colloquium with Gideon Simpson
Gideon Simpson from Drexel University will be giving a talk titled, "Sampling from Rough Energy Landscapes."
Abstract:
In high dimensional problems, classical method for numerical integration become computationally inefficient, leading us to consider the Monte Carlo method, an algorithm based on sampling from an appropriately chosen probability distribution. In physics and chemistry, such high dimensional problems readily occur when one examines the interactions of a large number of atoms interacting through an energy landscape at finite temperature; these are the so called Boltzmann distributions.
In this talk, I will introduce Monte Carlo sampling for such distributions, and examine distinct challenges that appear when the landscape is, in some sense, “rough.” We develop and examine sampling strategies that are insensitive to this roughness, and analyze how the roughness can impede the performance of other sampling strategies. Numerical experiments confirm our predictions. This is joint work with Petr Plechac (U. Delaware).
Bryn Mawr College welcomes the full participation of all individuals in all aspects of campus life. Should you wish to request a disability-related accommodation for this event, please contact the event sponsor/coordinator. Requests should be made as early as possible.