Showing 1–6 of 6 results for author: Syvari, M

Revealing systematics in phenomenologically viable flux vacua with reinforcement learning

Authors: Sven Krippendorf, Rene Kroepsch, Marc Syvaeri

Abstract: The organising principles underlying the structure of phenomenologically viable string vacua can be accessed by sampling such vacua. In many cases this is prohibited by the computational cost of standard sampling methods in the high dimensional model space. Here we show how this problem can be alleviated using reinforcement learning techniques to explore string flux vacua. We demonstrate in the ca… ▽ More The organising principles underlying the structure of phenomenologically viable string vacua can be accessed by sampling such vacua. In many cases this is prohibited by the computational cost of standard sampling methods in the high dimensional model space. Here we show how this problem can be alleviated using reinforcement learning techniques to explore string flux vacua. We demonstrate in the case of the type IIB flux landscape that vacua with requirements on the expectation value of the superpotential and the string coupling can be sampled significantly faster by using reinforcement learning than by using metropolis or random sampling. Our analysis is on conifold and symmetric torus background geometries. We show that reinforcement learning is able to exploit successful strategies for identifying such phenomenologically interesting vacua. The strategies are interpretable and reveal previously unknown correlations in the flux landscape. △ Less

Submitted 8 July, 2021; originally announced July 2021.

Comments: 30 pages, 12 figures, 2 tables

Report number: LMU-ASC 20/21, MPP-2021-108

Improving Simulations with Symmetry Control Neural Networks

Authors: Marc Syvaeri, Sven Krippendorf

Abstract: The dynamics of physical systems is often constrained to lower dimensional sub-spaces due to the presence of conserved quantities. Here we propose a method to learn and exploit such symmetry constraints building upon Hamiltonian Neural Networks. By enforcing cyclic coordinates with appropriate loss functions, we find that we can achieve improved accuracy on simple classical dynamics tasks. By fitt… ▽ More The dynamics of physical systems is often constrained to lower dimensional sub-spaces due to the presence of conserved quantities. Here we propose a method to learn and exploit such symmetry constraints building upon Hamiltonian Neural Networks. By enforcing cyclic coordinates with appropriate loss functions, we find that we can achieve improved accuracy on simple classical dynamics tasks. By fitting analytic formulae to the latent variables in our network we recover that our networks are utilizing conserved quantities such as (angular) momentum. △ Less

Submitted 29 April, 2021; originally announced April 2021.

Comments: 7 pages, 2 figures, accepted as workshop paper at ICLR 2021 SimDL Workshop

Report number: LMU-ASC 11/21, MPP-2021-67

Integrability ex machina

Authors: Sven Krippendorf, Dieter Lust, Marc Syvaeri

Abstract: Determining whether a dynamical system is integrable is generally a difficult task which is currently done on a case by case basis requiring large human input. Here we propose and test an automated method to search for the existence of relevant structures, the Lax pair and Lax connection respectively. By formulating this search as an optimization problem, we are able to identify appropriate struct… ▽ More Determining whether a dynamical system is integrable is generally a difficult task which is currently done on a case by case basis requiring large human input. Here we propose and test an automated method to search for the existence of relevant structures, the Lax pair and Lax connection respectively. By formulating this search as an optimization problem, we are able to identify appropriate structures via machine learning techniques. We test our method on standard systems of classical integrability and find that we can single out some integrable deformations of a system. Due to the ambiguity in defining a Lax pair our algorithm identifies novel Lax pairs which can be easily verified analytically. △ Less

Submitted 29 April, 2021; v1 submitted 12 March, 2021; originally announced March 2021.

Comments: 18 pages, 1 figure, v2: modified integrable perturbation discussion

Report number: LMU-ASC 05/21, MPP-2021-29

Detecting Symmetries with Neural Networks

Authors: Sven Krippendorf, Marc Syvaeri

Abstract: Identifying symmetries in data sets is generally difficult, but knowledge about them is crucial for efficient data handling. Here we present a method how neural networks can be used to identify symmetries. We make extensive use of the structure in the embedding layer of the neural network which allows us to identify whether a symmetry is present and to identify orbits of the symmetry in the input.… ▽ More Identifying symmetries in data sets is generally difficult, but knowledge about them is crucial for efficient data handling. Here we present a method how neural networks can be used to identify symmetries. We make extensive use of the structure in the embedding layer of the neural network which allows us to identify whether a symmetry is present and to identify orbits of the symmetry in the input. To determine which continuous or discrete symmetry group is present we analyse the invariant orbits in the input. We present examples based on rotation groups $SO(n)$ and the unitary group $SU(2).$ Further we find that this method is useful for the classification of complete intersection Calabi-Yau manifolds where it is crucial to identify discrete symmetries on the input space. For this example we present a novel data representation in terms of graphs. △ Less

Submitted 30 March, 2020; originally announced March 2020.

Comments: 20 pages, 9 figures

Report number: LMU-ASC 11/20, MPP-2020-34

Open-String Non-Associativity in an R-flux Background

Authors: Dieter Lust, Emanuel Malek, Erik Plauschinn, Marc Syvari

Abstract: We derive the commutation relations for open-string coordinates on D-branes in non-geometric background spaces. Starting from D0-branes on a three-dimensional torus with H-flux, we show that open strings with end points on D3-branes in a three-dimensional R-flux background exhibit a non-associative phase-space algebra, which is similar to the non-associative R-flux algebra of closed strings. There… ▽ More We derive the commutation relations for open-string coordinates on D-branes in non-geometric background spaces. Starting from D0-branes on a three-dimensional torus with H-flux, we show that open strings with end points on D3-branes in a three-dimensional R-flux background exhibit a non-associative phase-space algebra, which is similar to the non-associative R-flux algebra of closed strings. Therefore, the effective open-string gauge theory on the D3-branes is expected to be a non-associative gauge theory. We also point out differences between the non-associative phase space structure of open and closed strings in non-geometric backgrounds, which are related to the different structure of the world-sheet commutators of open and closed strings. △ Less

Submitted 5 June, 2020; v1 submitted 13 March, 2019; originally announced March 2019.

Comments: 29 pages; v2: added section 3.5, minor changes; v3: published version, clarifications added in sections 3 and 4, added appendix B.4

Locally non-geometric fluxes and missing momenta in M-theory

Authors: Dieter Lust, Emanuel Malek, Marc Syvari

Abstract: We use exceptional field theory to describe locally non-geometric spaces of M-theory of more than three dimensions. For these spaces, we find a new set of locally non-geometric fluxes which lie in the R-R sector in the weak-coupling limit and can typically be characterised by mixed symmetry tensors. These spaces thus provide new examples of non-geometric backgrounds which go beyond the NS-NS secto… ▽ More We use exceptional field theory to describe locally non-geometric spaces of M-theory of more than three dimensions. For these spaces, we find a new set of locally non-geometric fluxes which lie in the R-R sector in the weak-coupling limit and can typically be characterised by mixed symmetry tensors. These spaces thus provide new examples of non-geometric backgrounds which go beyond the NS-NS sector of string theory. Starting from twisted tori we construct duality chains that lead to these new non-geometric backgrounds and we show that, just as in the four-dimensional case, there are missing momenta associated to the mixed symmetry tensors. △ Less

Submitted 12 January, 2018; v1 submitted 16 October, 2017; originally announced October 2017.

Comments: 49 pages; v2: Published version, minor changes made

Report number: LMU-ASC 63/17, MPP-2017-223

Journal ref: JHEP 1801 (2018) 50