Speaker:
Jim Halverson
Institution:
Northeastern University
Date:
Tuesday, October 9, 2018
Time:
2:00 pm
Location:
FRH 4135
*Special Day / Time / Location
Abstract:
In this talk I will exemplify how a variety of techniques from data science can be used to better understand the string landscape and its remnant degrees of freedom. By "string remnants," I mean degrees of freedom that often arise in concrete string constructions, but are not present for the purpose of solving a low-energy problem. They are remnants of the ultraviolet construction. Nevertheless, these remnant degrees of freedom can be relevant for particle physics and cosmology, including providing constraints on string constructions. Specifically, I will present a concrete ensemble of 10^755 topologically distinct F-theory geometries that form a connected Calabi-Yau moduli space, and study it using a variety of techniques from data science. Large gauge sectors and strong string coupling are both generic. Features related to E6 GUTs will be understood via supervised machine learning and conjecture generation. Deep reinforcement learning will be used to probe the boundary between weakly coupled and strongly coupled theories. A dark matter oversaturation problem in the landscape will be highlighted, that is perhaps akin to the cosmological moduli problem. The goal of the talk is to demonstrate ways in which we might learn particle-cosmology lessons from the landscape, and how data science may play a role.
In this talk I will exemplify how a variety of techniques from data science can be used to better understand the string landscape and its remnant degrees of freedom. By "string remnants," I mean degrees of freedom that often arise in concrete string constructions, but are not present for the purpose of solving a low-energy problem. They are remnants of the ultraviolet construction. Nevertheless, these remnant degrees of freedom can be relevant for particle physics and cosmology, including providing constraints on string constructions. Specifically, I will present a concrete ensemble of 10^755 topologically distinct F-theory geometries that form a connected Calabi-Yau moduli space, and study it using a variety of techniques from data science. Large gauge sectors and strong string coupling are both generic. Features related to E6 GUTs will be understood via supervised machine learning and conjecture generation. Deep reinforcement learning will be used to probe the boundary between weakly coupled and strongly coupled theories. A dark matter oversaturation problem in the landscape will be highlighted, that is perhaps akin to the cosmological moduli problem. The goal of the talk is to demonstrate ways in which we might learn particle-cosmology lessons from the landscape, and how data science may play a role.
Host:
Michael Ratz