Conference Agenda

Overview and details of the sessions of this conference. Please select a date or location to show only sessions at that day or location. Please select a single session for detailed view (with abstracts and downloads if available).

 
Session Overview
Session
MS193: Algebraic geometry, data science and fundamental physics
Time:
Friday, 12/Jul/2019:
3:00pm - 5:00pm

Location: Unitobler, F-111
30 seats, 56m^2

Presentations
3:00pm - 5:00pm

Algebraic geometry, data science and fundamental physics

Chair(s): Yang-Hui He (City, University of London, Oxford University & Nankai), Fabian Ruehle (CERN & Oxford University), Heather Harrington (Oxford University)

There has been an increasing interaction between computational algebraic geometry, data science and fundamental theoretical physics.

This is rooted in the tradition that the 2 pillars of theoretical physics- general relativity and the standard model of particle physics, as well as their best candiate unified theory of superstrings - are physical realizations of the study of gauge connections and Riemannian metrics on manifolds.

In the last couple of years, problems such as mapping the Calabi-Yau landscape, translating problems in particle theory to precise problems in algebraic and differential geometry, using the latest techniques in machine-learning, etc., have taken off in the theoretical physics community.

This session in SIAM AG 2019 is a perfect venue for further explorations.

 

(25 minutes for each presentation, including questions, followed by a 5-minute break; in case of x<4 talks, the first x slots are used unless indicated otherwise)

 

The Calabi-Yau landscape & machine learning

Yang-Hui He
City, University of London, Oxford University & Nankai

To be completed.

 

Machine Learning for String Vacua

Fabian Ruehle
CERN & Oxford University

I will discuss complexity classes of problems encountered in string theory. Since most problems are NP-complete or undecidable, data science techniques are used to tackle them. As an example, I will present Reinforcement Learning applications to string landscape questions and demonstrate how the algorithm learns to solve the associated problem.

 

Knot Theory and Machine Learning

Jim Halverson
Northeastern

I will discuss various aspects of studying knot theory and knot topological invariants with machine learning.

 

Machine-learning a virus assembly fitness landscape

Pierre-Philippe Dechant
York St John

Realistic evolutionary fitness landscapes are notoriously difficult to construct. A recent cutting-edge model of virus assembly is based on a detailed understanding of the geometry involved and fundamental biophysical principles, which allows one to capture the contribution to fitness coming from assembly efficiency in a suitably quantitative way. This model has a virus capsid shell consisting of twelve pentagons in a dodecahedral arrangement. Furthermore, there are 12 corresponding packaging signals - features in the genome which help recruit the twelve pentagonal capsid building blocks onto the growing capsid - in three binding affinity bands. The complete assembly phenotype space consisting of 312genomes has been explored via computationally expensive stochastic ab initio assembly models on a supercomputer, giving a fitness landscape in terms of the assembly efficiency.

Using machine-learning techniques, we have shown that the intensive computation can be short-circuited in a matter of minutes to astounding accuracy. There is thus a large hidden degeneracy in the detailed microphysical models, which allows one to understand general features that emerge at a higher level. This opens up the possibility of tackling more complicated models by bootstrapping, i.e. by only partially exploring the phenotype space in order to machine learn the generic features and then to use these to predict the remainder of the fitness landscape.