Dr. William J. Cunningham
Will Cunningham
Postdoctoral Fellow
Perimeter Institute
Biography
Will Cunningham is a postdoctoral fellow at Perimeter Institute for Theoretical Physics. He works in the Quantum Gravity group on the Discretuum to Continuum Initiative.
After studying Lattice Quantum Chromodynamics during his undergraduate studies at Rensselaer Polytechnic Institute, he began work on Network Science and Causal Set Quantum Gravity during his Ph.D. at Northeastern University. While in graduate school, he specialized in computer microarchitecture and high performance algorithms while developing new software for the broader community. He is currently working on projects involving the statistical physics of random graphs and deep learning in quantum gravity.
Academic
Perimeter Institute for Theoretical Physics (2018-2021)
Postdoctoral Fellow
Research Focus: Discretuum to Continuum Initiative
Northeastern University (2013-2018)
Ph.D. Physics, M.S. Physics
Dissertation: High Performance Algorithms for Quantum Gravity and Cosmology
Rensselaer Polytechnic Institute (2009-2013)
B.S. Physics
Concentrations: Lattice QCD, Astrophysics
George Mason High School (2005-2009)
International Baccalaureate Diploma
Concentration: Chemistry
Research
Statistical Physics of Causal Sets
"We are currently studying the ensemble of 2D causal sets by constructing Markov chains of causal sets with variable geometry and topology. The partition function is furnished by the 2D Benincasa-Dowker action, which is known to approximate the Einstein-Hilbert action in the large-N limit. During this process we will better characterize the phases of 2D causal sets, with future extensions to higher dimension."
Machine Learning in Quantum Gravity
"Is machine learning a useful tool in quantum gravity? We claim it is, particularly for classification tasks. A simple proof-of-concept experiment showed a deep neural network can be trained to classify the dimension and manifold of causal sets sprinkled into various spacetimes. In future work, we will use deep reinforcement learning to learn how causal sets grow while extremizing the action."
Boundaries in Causal Sets
"What does a boundary look like in a discrete, non-local object like a causal set? Recent work showed causal sets can be partitioned in several ways which gives a characterization of the convex hull, and we developed an algorithm to measure the volume of timelike boundaries. Yet it remains unknown what the timelike boundary term for the causal set action will look like. To better understand boundaries, we are working to calculate the causal set equivalent of the Gibbons-Hawking-York terms from general relativity."
Vacuum Selection in F Theory
"F theory details a string landscape of at least 10755 geometries, each of which can support at least 10500 flux vacua. Somewhere within this landscape lies the Standard Model of Physics, yet there exist no good methods to fully explore such a vast space. We take a first step at characterizing the string landscape using a network model, where nodes are geometries and links represent simple topological transitions. This model is the first to give evidence that at late times in a bubble cosmology, there exists a heterogenous distribution of selected vacua."
Navigation in Random Geometric Graphs
"Geometric graphs in hyperbolic spaces explain the optimality of many network functions related to finding paths in the network without global knowledge of the network structure. Yet little is known about routing behavior in non-Riemannian latent spaces. We study the navigability of undirected random geometric graphs in three Lorentzian manifolds: the de Sitter spacetime, the Einstein-de Sitter spacetime, and an interpolating spacetime similar to our physical universe. We find graphs in such spacetimes are navigable only when dark energy is present."
Talks
Inference of Boundaries in Causal Sets — U. Heidelberg and NetSci2018 (06/2018)
A new algorithm identifies and measures timelike boundaries in (1+1)-dimensional Minkowski causal sets. Slides
Deep Learning in Quantum Gravity — Quantum Gravity on the Computer (03/2018)
Deep supervised learning can be used to classify causal sets by manifold and dimension. Slides
Vacuum Selection from Cosmology Using Networks of String Geometries — Rensselaer Polytechnic Institute (01/2018)
A network model of string geometries shows evidence of vacuum selection in a late-time bubble cosmology. Slides
The Big Data Approach to Quantum Gravity — Perimeter Institute (12/2017)
Efficient parallel algorithms allow us to better study large problems in causal set theory and F theory. Slides Video
Introduction to Network Science — String Data Workshop (11/2017)
Network science is useful for modeling many complex systems we observe in the world. Slides
Timelike Boundary Terms in the Causal Set Action — Raman Research Institute (12/2016)
Timelike boundaries can be identified in causal sets, and perhaps measured. Slides
An Introduction to Parallel Programming: OpenMP, SSE/AVX, and MPI — Northeastern U. (04/2016)
Parallel computational techniques are easy to incorporate into graph problems. Slides
Code
Causal Set Generator
Generate and analyze causal sets and other random geometric graphs using ultra-efficient parallel techniques. Learn more on Bitbucket.
FastMath Toolkit
This toolkit provides optimized mathematical functions and compact data structures along with other useful utilities. Learn more on Bitbucket.