arXiv subjects

Sven Krippendorf

Publications and source records attributed to Sven Krippendorf.

At least 37 records · Page 2Linked to original sources

arXiv2023

The eROSITA Final Equatorial-Depth Survey (eFEDS): A Machine Learning Approach to Infer Galaxy Cluster Masses from eROSITA X-ray Images

We develop a neural network based pipeline to estimate masses of galaxy clusters with a known redshift directly from photon information in X-rays. Our neural networks are trained using supervised learning on simulations of eROSITA observations, focusing in this paper on the Final Equatorial Depth Survey (eFEDS). We use convolutional neural networks which are modified to include additional information of the cluster, in particular its redshift. In contrast to existing work, we utilize simulations including background and point sources to develop a tool which is usable directly on observational eROSITA data for an extended mass range from group size halos to massive clusters with masses in between $10^{13}M_\odot<M<10^{15}M_\odot.$ Using this method, we are able to provide for the first time neural network mass estimation for the observed eFEDS cluster sample from Spectrum-Roentgen-Gamma/eROSITA observations and we find consistent performance with weak lensing calibrated masses. In this measurement, we do not use weak lensing information and we only use previous cluster mass information which was used to calibrate the cluster properties in the simulations. When compared to simulated data, we observe a reduced scatter with respect to luminosity and count-rate based scaling relations. We comment on the application for other upcoming eROSITA All-Sky Survey observations.

astro-ph.CO↗

arXiv2022

Towards structures in the flux landscape at large number of moduli

Sampling string flux vacua enables us to study structures in the string landscape. Here we demonstrate that sampling at large number of moduli is possible for the simplified landscape model of ADK. Using dimensional reduction, we identify analytic structures in these samples. In this example, we find that these structures are rather insensitive to the underlying distribution of UV parameters but they emerge only at large number of moduli and they can be attributed to sampling bias.

hep-th↗

arXiv2022

Updated Bounds on Axion-Like Particles from X-ray Observations

In this work we revisit five different point sources within or behind galaxy clusters in order to constrain the coupling constant between axion-like particles (ALPs) and photons. We use three distinct machine learning (ML) techniques and compare our results with a standard $χ^2$ analysis. For the first time we apply approximate Bayesian computation to searches for ALPs and find consistently good performance across ML classifiers. Further, we apply more realistic 3D magnetic field simulations of galaxy clusters and compare our results with previously used 1D simulations. We find constraints on the ALP-photon coupling at the level of state-of-the-art bounds with $g_{aγγ} \lesssim 0.6 \times 10^{-12}$ GeV${}^{-1}$, hence improving on previous constraints obtained from the same observations.

astro-ph.CO↗

arXiv2022

A duality connecting neural network and cosmological dynamics

We demonstrate that the dynamics of neural networks trained with gradient descent and the dynamics of scalar fields in a flat, vacuum energy dominated Universe are structurally profoundly related. This duality provides the framework for synergies between these systems, to understand and explain neural network dynamics and new ways of simulating and describing early Universe models. Working in the continuous-time limit of neural networks, we analytically match the dynamics of the mean background and the dynamics of small perturbations around the mean field, highlighting potential differences in separate limits. We perform empirical tests of this analytic description and quantitatively show the dependence of the effective field theory parameters on hyperparameters of the neural network. As a result of this duality, the cosmological constant is matched inversely to the learning rate in the gradient descent update.

gr-qc↗

arXiv2021

Probing the Structure of String Theory Vacua with Genetic Algorithms and Reinforcement Learning

Identifying string theory vacua with desired physical properties at low energies requires searching through high-dimensional solution spaces - collectively referred to as the string landscape. We highlight that this search problem is amenable to reinforcement learning and genetic algorithms. In the context of flux vacua, we are able to reveal novel features (suggesting previously unidentified symmetries) in the string theory solutions required for properties such as the string coupling. In order to identify these features robustly, we combine results from both search methods, which we argue is imperative for reducing sampling bias.

hep-th↗

arXiv2021

Revealing systematics in phenomenologically viable flux vacua with reinforcement learning

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.

hep-th↗

arXiv2021

Moduli-dependent Calabi-Yau and SU(3)-structure metrics from Machine Learning

We use machine learning to approximate Calabi-Yau and SU(3)-structure metrics, including for the first time complex structure moduli dependence. Our new methods furthermore improve existing numerical approximations in terms of accuracy and speed. Knowing these metrics has numerous applications, ranging from computations of crucial aspects of the effective field theory of string compactifications such as the canonical normalizations for Yukawa couplings, and the massive string spectrum which plays a crucial role in swampland conjectures, to mirror symmetry and the SYZ conjecture. In the case of SU(3) structure, our machine learning approach allows us to engineer metrics with certain torsion properties. Our methods are demonstrated for Calabi-Yau and SU(3)-structure manifolds based on a one-parameter family of quintic hypersurfaces in $\mathbb{P}^4.$

hep-th↗

arXiv2021

Improving Simulations with Symmetry Control Neural Networks

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.

cs.LG↗

arXiv2021

Integrability ex machina

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.

nlin.SI↗

arXiv2020

UV Shadows in EFTs: Accidental Symmetries, Robustness and No-Scale Supergravity

We argue that accidental approximate scaling symmetries are robust predictions of weakly coupled string vacua, and show that their interplay with supersymmetry and other (generalised) internal symmetries underlies the ubiquitous appearance of no-scale supergravities in low-energy 4D EFTs. We identify 4 nested types of no-scale supergravities, and show how leading quantum corrections can break scale invariance while preserving some no-scale properties (including non-supersymmetric flat directions). We use these ideas to classify corrections to the low-energy 4D supergravity action in perturbative 10D string vacua, including both bulk and brane contributions. Our prediction for the Kähler potential at any fixed order in $α'$ and string loops agrees with all extant calculations. p-form fields play two important roles: they spawn many (generalised) shift symmetries; and space-filling 4-forms teach 4D physics about higher-dimensional phenomena like flux quantisation. We argue that these robust symmetry arguments suffice to understand obstructions to finding classical de Sitter vacua, and suggest how to get around them in UV complete models.

hep-th↗

arXiv2020

Detecting Symmetries with Neural Networks

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.

physics.comp-ph↗

arXiv2020

Accelerating the search for Axion-Like Particles with machine learning

Machine learning (ML) techniques have been applied with tremendous success in many areas of physics. In this work, we use ML to place bounds on the coupling between photons and axion-like particles (ALPs). This coupling causes ALPs and photons to interconvert in the presence of a background magnetic field. This would lead to modulations in the spectra of point sources shining through the magnetic fields of galaxy clusters. This effect has already been used to place world-leading bounds on the ALP-photon coupling using conventional statistical methods. We train ML classification algorithms on simulated spectra from the Chandra X-ray telescope for a range of point sources and ALP-photon couplings. We then use the response of these algorithms to the real Chandra spectra to place bounds on ALP-photon interactions. We obtain bounds at a similar level to those based on other techniques, but find improvements on an individual source basis. We expect such search techniques to become increasingly important for ALP searches with future telescopes that will offer substantially higher energy resolution.

astro-ph.HE↗

arXiv2020

Connecting Dualities and Machine Learning

Dualities are widely used in quantum field theories and string theory to obtain correlation functions at high accuracy. Here we present examples where dual data representations are useful in supervised classification, linking machine learning and typical tasks in theoretical physics. We then discuss how such beneficial representations can be enforced in the latent dimension of neural networks. We find that additional contributions to the loss based on feature separation, feature matching with respect to desired representations, and a good performance on a `simple' correlation function can lead to known and unknown dual representations. This is the first proof of concept that computers can find dualities. We discuss how our examples, based on discrete Fourier transformation and Ising models, connect to other dualities in theoretical physics, for instance Seiberg duality.

physics.comp-ph↗

arXiv2018

GANs for generating EFT models

We initiate a way of generating models by the computer, satisfying both experimental and theoretical constraints. In particular, we present a framework which allows the generation of effective field theories. We use Generative Adversarial Networks to generate these models and we generate examples which go beyond the examples known to the machine. As a starting point, we apply this idea to the generation of supersymmetric field theories. In this case, the machine knows consistent examples of supersymmetric field theories with a single field and generates new examples of such theories. In the generated potentials we find distinct properties, here the number of minima in the scalar potential, with values not found in the training data. We comment on potential further applications of this framework.

cs.LG↗

arXiv2018

Moduli Stars

We explore the possibility that (Bose-Einstein) condensation of scalar fields from string compactifications can lead to long-lived compact objects. Depending on the type of scalar fields we find different realisations of star-like and solitonic objects. We illustrate in the framework of type~IIB string compactifications that closed string moduli can lead to heavy microscopic stars with masses of order $\mathcal{V}^αM_{\rm Planck}$, $α=1,3/2,5/3$ where $\mathcal{V}$ is the volume of the extra dimensions. Macroscopic compact objects from ultra-light string axions are realised with masses of order $e^{\mathcal{V}^{2/3}}M_{\rm Planck}.$ Q-ball configurations can be obtained from open string moduli whereas the closed string sector gives rise to a new class of solutions, named PQ-balls, that arise in the two-field axion-modulus system. The stability, the potential for the formation, and the observability of moduli stars through gravitational waves are discussed. In particular we point out that during the early matter phase given by moduli domination, density perturbations grow by a factor $\mathcal{V}^β$ with $β>2$ and non-linear effects cannot be neglected.

hep-th↗

arXiv2018

Searching for Axion-Like Particles with X-ray Polarimeters

X-ray telescopes are an exceptional tool for searching for new fundamental physics. In particular, X-ray observations have already placed world-leading bounds on the interaction between photons and axion-like particles (ALPs). ALPs are hypothetical new ultra-light particles motivated by string theory models. They can also act as dark matter and dark energy, and provide a solution to the strong CP problem. In a background magnetic field, ALPs and photons may interconvert. This leads to energy dependent modulations in both the flux and polarisation of the spectra of point sources shining through large magnetic fields. The next generation of polarising X-ray telescopes will offer new detection possibilities for ALPs. Here we present techniques and projected bounds for searching for ALPs with X-ray polarimetry. We demonstrate that upcoming X-ray polarimetry missions have the potential to place world-leading bounds on ALPs.

hep-ph↗

arXiv2017

Projected bounds on ALPs from Athena

Galaxy clusters represent excellent laboratories to search for Axion-Like Particles (ALPs). They contain magnetic fields which can induce quasi-sinusoidal oscillations in the X-ray spectra of AGNs situated in or behind them. Due to its excellent energy resolution, the X-ray Integral Field Unit (X-IFU) instrument onboard the Athena X-ray Observatory will be far more sensitive to ALP-induced modulations than current detectors. As a first analysis of the sensitivity of Athena to the ALP-photon coupling $g_{a γγ}$, we simulate observations of the Seyfert galaxy NGC1275 in the Perseus cluster using the SIXTE simulation software. We estimate that for a 200ks exposure, a non-observation of spectral modulations will constrain ${g_{aγγ}\lesssim1.5\times10^{-13}\rm{GeV}^{-1}}$ for $m_a \lesssim 10^{-12} \rm{eV}$, representing an order of magnitude improvement over constraints derived using the current generation of satellites.

astro-ph.HE↗

arXiv2017

Consistency of Hitomi, XMM-Newton and Chandra 3.5 keV data from Perseus

\emph{Hitomi} observations of Perseus with the Soft X-ray Spectrometer (SXS) provide a high-resolution look at the 3.5 keV feature reported by multiple groups in the Perseus cluster. The \emph{Hitomi} spectrum -- which involves the sum of diffuse cluster emission and the point-like central Active Galactic Nucleus (AGN) -- does not show any excess at $E \sim 3.5 {\rm keV}$, giving an apparent inconsistency with previous observations of excess diffuse emission. We point out that 2009 \emph{Chandra} data reveals a strong dip in the AGN spectrum at $E = (3.54 \pm 0.02) {\rm keV}$ (cluster frame) -- the identical energy to the diffuse excess observed by \emph{XMM-Newton}. Scaling this dip to the 2016 AGN luminosity and adding it to the diffuse \emph{XMM-Newton} excess, this predicts an overall dip in the SXS field of view of $(-5.9 \pm 4.4) \times 10^{-6} \, {\rm ph} \, {\rm cm}^{-2} \, {\rm s}^{-1}$ at $E= 3.54$ keV -- a precise match to the \emph{Hitomi} data when broadened by the dark matter virial velocity. We describe models of Fluorescent Dark Matter that can reproduce this physics, in which dark matter absorbs and then re-emits 3.5 keV photons emitted from the central AGN.

astro-ph.HE↗