Filter and export

Publications

What is a Publication?
8 Publications visible to you, out of a total of 8
Gravitational-wave model for neutron star merger remnants with supervised learning

Abstract (Expand)

We present a time-domain model for the gravitational waves emitted by equal-mass binary neutron star merger remnants for a fixed equation of state. We construct a large set of numerical relativity … simulations for a single equation of state consistent with current constraints, totaling 157 equal-mass binary neutron star merger configurations. The gravitational-wave model is constructed using the supervised learning method of K-nearest neighbor regression. As a first step toward developing a general model with supervised learning methods that accounts for the dependencies on equation of state and the binary masses of the system, we explore the impact of the size of the dataset on the model. We assess the accuracy of the model for a varied dataset size and number density in total binary mass. Specifically, we consider five training sets of simulations uniformly distributed in total binary mass. We evaluate the resulting models in terms of faithfulness using a test set of 30 additional simulations that are not used during training and which are equidistantly spaced in total binary mass. The models achieve faithfulness with maximum values in the range of 0.980 to 0.995. We assess our models simulating signals observed by the three-detector network of Advanced LIGO-Virgo. We find that all models with training sets of size equal to or larger than 40 achieve an unbiased measurement of the main gravitational-wave frequency. We confirm that our results do not depend qualitatively on the choice of the (fixed) equation of state. We conclude that training sets, with a minimum size of 40 simulations, or a number density of approximately 11 simulations per 0.1⁢𝑀⊙ of total binary mass, suffice for the construction of faithful templates for the postmerger signal for a single equation of state and equal-mass binaries, and lead to mean faithfulness values of ℱ ≃0.95. Our model being based on only one fixed equation of state represents only a first step toward a method that is fully applicable for gravitational-wave parameter estimation. However, our findings are encouraging since we show that our supervised learning model built on a set of simulations for a fixed equation of state successfully recovers the main gravitational-wave features of a simulated signal obtained using another equation of state. This may indicate that the extension of this model to an arbitrary equation of state may actually be achieved with a manageable set of simulations.

Authors: Theodoros Soultanis, Kiril Maltsev, Andreas Bauswein, Katerina Chatziioannou, Friedrich K. Röpke, Nikolaos Stergioulas

Date Published: 2025

Publication Type: Journal

Substituting density functional theory in reaction barrier calculations for hydrogen atom transfer in proteins

Abstract (Expand)

Hydrogen atom transfer (HAT) reactions are important in many biological systems. As these reactions are hard to observe experimentally, it is of high interest to shed light on them using simulations. … Here, we present a machine learning model based on graph neural networks for the prediction of energy barriers of HAT reactions in proteins. As input, the model uses exclusively non-optimized structures as obtained from classical simulations. It was trained on more than 17 000 energy barriers calculated using hybrid density functional theory. We built and evaluated the model in the context of HAT in collagen, but we show that the same workflow can easily be applied to HAT reactions in other biological or synthetic polymers. We obtain for relevant reactions (small reaction distances) a model with good predictive power (R2 ∼ 0.9 and mean absolute error of <3 kcal mol−1). As the inference speed is high, this model enables evaluations of dozens of chemical situations within seconds. When combined with molecular dynamics in a kinetic Monte-Carlo scheme, the model paves the way toward reactive simulations.

Authors: Kai Riedmiller, Patrick Reiser, Elizaveta Bobkova, Kiril Maltsev, Ganna Gryn’ova, Pascal Friederich, Frauke Gräter

Date Published: 14th Feb 2024

Publication Type: Journal

Scalable stellar evolution forecasting — Deep learning emulation versus hierarchical nearest-neighbor interpolation

Abstract

Not specified

Authors: K. Maltsev, F. R. N. Schneider, F. K. Röpke, A. I. Jordan, G. A. Qadir, W. E. Kerzendorf, K. Riedmiller, P. van der Smagt

Date Published: 19th Jan 2024

Publication Type: Journal

Substituting density functional theory in reaction barrier calculations for hydrogen atom transfer in proteins

Abstract (Expand)

Hydrogen atom transfer (HAT) reactions are important in many biological systems. As these reactions are hard to observe experimentally, it is of high interest to shed light on them using simulations. … Here, we present a machine learning model based on graph neural networks for the prediction of energy barriers of HAT reactions in proteins. As input, the model uses exclusively non-optimized structures as obtained from classical simulations. It was trained on more than 17 000 energy barriers calculated using hybrid density functional theory. We built and evaluated the model in the context of HAT in collagen, but we show that the same workflow can easily be applied to HAT reactions in other biological or synthetic polymers. We obtain for relevant reactions (small reaction distances) a model with good predictive power (R2 ∼ 0.9 and mean absolute error of <3 kcal mol−1). As the inference speed is high, this model enables evaluations of dozens of chemical situations within seconds. When combined with molecular dynamics in a kinetic Monte-Carlo scheme, the model paves the way toward reactive simulations.

Authors: Kai Riedmiller, Patrick Reiser, Elizaveta Bobkova, Kiril Maltsev, Ganna Gryn'ova, Pascal Friederich, Frauke Gräter

Date Published: 16th Jan 2024

Publication Type: Journal

Substituting density functional theory in reaction barrier calculations for hydrogen atom transfer in proteins

Abstract (Expand)

Hydrogen atom transfer (HAT) reactions are important in many biological systems. As these reactions are hard to observe experimentally, it is of high interest to shed light on them using simulations. … Here, we present a machine learning model based on graph neural networks for the prediction of energy barriers of HAT reactions in proteins. As input, the model uses exclusively non-optimized structures as obtained from classical simulations. It was trained on more than 17 000 energy barriers calculated using hybrid density functional theory. We built and evaluated the model in the context of HAT in collagen, but we show that the same workflow can easily be applied to HAT reactions in other biological or synthetic polymers. We obtain for relevant reactions (small reaction distances) a model with good predictive power (R2 ∼ 0.9 and mean absolute error of <3 kcal mol−1). As the inference speed is high, this model enables evaluations of dozens of chemical situations within seconds. When combined with molecular dynamics in a kinetic Monte-Carlo scheme, the model paves the way toward reactive simulations.

Authors: Kai Riedmiller, Patrick Reiser, Elizaveta Bobkova, Kiril Maltsev, Ganna Gryn’ova, Pascal Friederich, Frauke Gräter

Date Published: 16th Jan 2024

Publication Type: Journal

Scalable stellar evolution forecasting - Deep learning emulation versus hierarchical nearest-neighbor interpolation

Abstract (Expand)

Many astrophysical applications require efficient yet reliable forecasts of stellar evolution tracks. One example is population synthesis, which generates forward predictions of models for comparison … with observations. The majority of state-of-the-art rapid population synthesis methods are based on analytic fitting formulae to stellar evolution tracks that are computationally cheap to sample statistically over a continuous parameter range. The computational costs of running detailed stellar evolution codes, such as MESA, over wide and densely sampled parameter grids are prohibitive, while stellar-age based interpolation in-between sparsely sampled grid points leads to intolerably large systematic prediction errors. In this work, we provide two solutions for automated interpolation methods that offer satisfactory trade-off points between cost-efficiency and accuracy. We construct a timescale-adapted evolutionary coordinate and use it in a two-step interpolation scheme that traces the evolution of stars from zero age main sequence all the way to the end of core helium burning while covering a mass range from 0.65 to 300 M⊙. The feedforward neural network regression model (first solution) that we train to predict stellar surface variables can make millions of predictions, sufficiently accurate over the entire parameter space, within tens of seconds on a 4-core CPU. The hierarchical nearest-neighbor interpolation algorithm (second solution) that we hard-code to the same end achieves even higher predictive accuracy, the same algorithm remains applicable to all stellar variables evolved over time, but it is two orders of magnitude slower. Our methodological framework is demonstrated to work on the MESA Isochrones and Stellar Tracks (Choi et al. 2016) data set, but is independent of the input stellar catalog. Finally, we discuss the prospective applications of these methods and provide guidelines for generalizing them to higher dimensional parameter spaces.

Authors: K. Maltsev, F. R. N. Schneider, F. K. Röpke, A. I. Jordan, G. A. Qadir, W. E. Kerzendorf, K. Riedmiller, P. van der Smagt

Date Published: 2024

Publication Type: Journal

Stellar gravitational collapse, singularity formation and theory breakdown

Abstract (Expand)

We review three definitions (missing point(s) unsteadiness, infinite quadratic curvature invariant, and geodesic incompleteness) of what a gravitational singularity is, and argue that prediction of a …a gravitational singularity is problematic for General Relativity (GR), indicating breakdown of Lorentzian geometry, only insofar as it concerns the infinite curvature singularity characterization. In contrast, the geodesic incompleteness characterization is GR’s innovating hallmark, which is not meaningfully available in Newtonian gravity formulations (locally infinite density field, and locally infinite gravitational force) of what a gravitational singularity is. It is the continuous, non-quantized, nature of Lorentzian geometry which admits gravitational contraction be continued indefinitely. The Oppenheimer-Snyder 1939 analytical solution derives formation of a locally infinite curvature singularity and of incomplete geodesics, while Penrose’s 1965 theorem concerns formation of incomplete (null) geodesics only. We critically examine the main physical arguments against gravitational singularity formation in stellar collapse, with scope restriction to decades spanning in between Schwarzschild’s 1916 solution and Penrose’s 1965 singularity theorem. As the most robust curvature singularity formation counter-argument, we assess Markov’s derivation of an upper bound on the quadratic curvature invariant RμνλδRμνλδ≤1ℓP4 from a ratio of natural constants ħ, c and G, in connection with Wheeler’s grounding of the premise that the Planck scale ℓP is ultimate.

Author: Kiril Maltsev

Date Published: 1st Feb 2023

Publication Type: Journal

Powered by
 (v.1.16.0)
About HITS Publication SEEK | About HITS-Publication-SEEK | Funding and Programmes | Credits
Copyright © 2008 - 2024 The University of Manchester and HITS gGmbH