Searcharxiv⌕ Search

arXiv subjects

Sandro Wenzel

Publications and source records attributed to Sandro Wenzel.

17 recordsLinked to original sources
arXiv2023

Machine Learning methods for simulating particle response in the Zero Degree Calorimeter at the ALICE experiment, CERN

Currently, over half of the computing power at CERN GRID is used to run High Energy Physics simulations. The recent updates at the Large Hadron Collider (LHC) create the need for developing more efficient simulation methods. In particular, there exists a demand for a fast simulation of the neutron Zero Degree Calorimeter, where existing Monte Carlo-based methods impose a significant computational burden. We propose an alternative approach to the problem that leverages machine learning. Our solution utilises neural network classifiers and generative models to directly simulate the response of the calorimeter. In particular, we examine the performance of variational autoencoders and generative adversarial networks, expanding the GAN architecture by an additional regularisation network and a simple, yet effective postprocessing step. Our approach increases the simulation speed by 2 orders of magnitude while maintaining the high fidelity of the simulation.

cs.CV↗
arXiv2023

HEPScore: A new CPU benchmark for the WLCG

HEPScore is a new CPU benchmark created to replace the HEPSPEC06 benchmark that is currently used by the WLCG for procurement, computing resource pledges and performance studies. The development of the new benchmark, based on HEP applications or workloads, has involved many contributions from software developers, data analysts, experts of the experiments, representatives of several WLCG computing centres, as well as the WLCG HEPScore Deployment Task Force. In this contribution, we review the selection of workloads and the validation of the new HEPScore benchmark.

hep-ex↗
arXiv2022

Selectively increasing the diversity of GAN-generated samples

Generative Adversarial Networks (GANs) are powerful models able to synthesize data samples closely resembling the distribution of real data, yet the diversity of those generated samples is limited due to the so-called mode collapse phenomenon observed in GANs. Especially prone to mode collapse are conditional GANs, which tend to ignore the input noise vector and focus on the conditional information. Recent methods proposed to mitigate this limitation increase the diversity of generated samples, yet they reduce the performance of the models when similarity of samples is required. To address this shortcoming, we propose a novel method to selectively increase the diversity of GAN-generated samples. By adding a simple, yet effective regularization to the training loss function we encourage the generator to discover new data modes for inputs related to diverse outputs while generating consistent samples for the remaining ones. More precisely, we maximise the ratio of distances between generated images and input latent vectors scaling the effect according to the diversity of samples for a given conditional input. We show the superiority of our method in a synthetic benchmark as well as a real-life scenario of simulating data from the Zero Degree Calorimeter of ALICE experiment in LHC, CERN.

cs.CV↗
arXiv2020

End-to-end Sinkhorn Autoencoder with Noise Generator

In this work, we propose a novel end-to-end sinkhorn autoencoder with noise generator for efficient data collection simulation. Simulating processes that aim at collecting experimental data is crucial for multiple real-life applications, including nuclear medicine, astronomy and high energy physics. Contemporary methods, such as Monte Carlo algorithms, provide high-fidelity results at a price of high computational cost. Multiple attempts are taken to reduce this burden, e.g. using generative approaches based on Generative Adversarial Networks or Variational Autoencoders. Although such methods are much faster, they are often unstable in training and do not allow sampling from an entire data distribution. To address these shortcomings, we introduce a novel method dubbed end-to-end Sinkhorn Autoencoder, that leverages sinkhorn algorithm to explicitly align distribution of encoded real data examples and generated noise. More precisely, we extend autoencoder architecture by adding a deterministic neural network trained to map noise from a known distribution onto autoencoder latent space representing data distribution. We optimise the entire model jointly. Our method outperforms competing approaches on a challenging dataset of simulation data from Zero Degree Calorimeters of ALICE experiment in LHC. as well as standard benchmarks, such as MNIST and CelebA.

cs.LG↗
arXiv2013

Vectorising the detector geometry to optimize particle transport

Among the components contributing to particle transport, geometry navigation is an important consumer of CPU cycles. The tasks performed to get answers to "basic" queries such as locating a point within a geometry hierarchy or computing accurately the distance to the next boundary can become very computing intensive for complex detector setups. So far, the existing geometry algorithms employ mainly scalar optimisation strategies (voxelization, caching) to reduce their CPU consumption. In this paper, we would like to take a different approach and investigate how geometry navigation can benefit from the vector instruction set extensions that are one of the primary source of performance enhancements on current and future hardware. While on paper, this form of microparallelism promises increasing performance opportunities, applying this technology to the highly hierarchical and multiply branched geometry code is a difficult challenge. We refer to the current work done to vectorise an important part of the critical navigation algorithms in the ROOT geometry library. Starting from a short critical discussion about the programming model, we present the current status and first benchmark results of the vectorisation of some elementary geometry shape algorithms. On the path towards a full vector-based geometry navigator, we also investigate the performance benefits in connecting these elementary functions together to develop algorithms which are entirely based on the flow of vector-data. To this end, we discuss core components of a simple vector navigator that is tested and evaluated on a toy detector setup.

physics.comp-ph↗
arXiv2013

Zero-temperature Monte Carlo study of the non-coplanar phase of the classical bilinear-biquadratic Heisenberg model on the triangular lattice

We investigate the ground-state properties of the highly degenerate non-coplanar phase of the classical bilinear-biquadratic Heisenberg model on the triangular lattice with Monte Carlo simulations. For that purpose, we introduce an Ising pseudospin representation of the ground states, and we use a simple Metropolis algorithm with local updates, as well as a powerful cluster algorithm. At sizes that can be sampled with local updates, the presence of long-range order is surprisingly combined with an algebraic decay of correlations and the complete disordering of the chirality. It is only thanks to the investigation of unusually large systems (containing $\sim 10^8$ spins) with cluster updates that the true asymptotic regime can be reached and that the system can be proven to consist of equivalent (i.e., equally ordered) sublattices. These large-scale simulations also demonstrate that the scalar chirality exhibits long-range order at zero temperature, implying that the system has to undergo a finite-temperature phase transition. Finally, we show that the average distance in the order parameter space, which has the structure of an infinite Cayley tree, remains remarkably small between any pair of points, even in the limit when the real space distance between them tends to infinity.

cond-mat.str-el↗
arXiv2012

Evidence of columnar order in the fully frustrated transverse field Ising model on the square lattice

Using extensive classical and quantum Monte Carlo simulations, we investigate the ground-state phase diagram of the fully frustrated transverse field Ising model on the square lattice. We show that pure columnar order develops in the low-field phase above a surprisingly large length scale, below which an effective U(1) symmetry is present. The same conclusion applies to the Quantum Dimer Model with purely kinetic energy, to which the model reduces in the zero-field limit, as well as to the stacked classical version of the model. By contrast, the 2D classical version of the model is shown to develop plaquette order. Semiclassical arguments show that the transition from plaquette to columnar order is a consequence of quantum fluctuations.

cond-mat.str-el↗
arXiv2011

Monte Carlo study of the critical properties of the three-dimensional 120-degree model

We report on large scale finite-temperature Monte Carlo simulations of the classical $120^\circ$ or $e_g$ orbital-only model on the simple cubic lattice in three dimensions with a focus towards its critical properties. This model displays a continuous phase transition to an orbitally ordered phase. While the correlation length exponent $ν\approx0.665$ is close to the 3D XY value, the exponent $η\approx 0.15$ differs substantially from O(N) values. We also introduce a discrete variant of the $e_g$ model, called $e_g$-clock model, which is found to display the same set of exponents. Further, an emergent U(1) symmetry is found at the critical point $T_c$, which persists for $T<T_c$ below a crossover length scaling as $Λ\sim ξ^a$, with an unusually small $a\approx1.3$.

cond-mat.str-el↗
arXiv2011

Unveiling the nature of three dimensional orbital ordering transitions: the case of $e_g$ and $t_{2g}$ models on the cubic lattice

We perform large scale finite-temperature Monte Carlo simulations of the classical $e_g$ and $t_{2g}$ orbital models on the simple cubic lattice in three dimensions. The $e_g$ model displays a continuous phase transition to an orbitally ordered phase. While the correlation length exponent $ν\approx0.66(1)$ is close to the 3D XY value, the exponent $η\approx 0.15(1)$ differs substantially from O(N) values. At $T_c$ a U(1) symmetry emerges, which persists for $T<T_c$ below a crossover length scaling as $Λ\sim ξ^a$, with an unusually small $a\approx1.3$. Finally, for the $t_{2g}$ model we find a {\em first order} transition into a low-temperature lattice-nematic phase without orbital order.

cond-mat.str-el↗
arXiv2010

Re-examining the directional-ordering transition in the compass model with screw-periodic boundary conditions

We study the directional-ordering transition in the two-dimensional classical and quantum compass models on the square lattice by means of Monte Carlo simulations. An improved algorithm is presented which builds on the Wolff cluster algorithm in one-dimensional subspaces of the configuration space. This improvement allows us to study classical systems up to $L=512$. Based on the new algorithm we give evidence for the presence of strongly anomalous scaling for periodic boundary conditions which is much worse than anticipated before. We propose and study alternative boundary conditions for the compass model which do not make use of extended configuration spaces and show that they completely remove the problem with finite-size scaling. In the last part, we apply these boundary conditions to the quantum problem and present a considerably improved estimate for the critical temperature which should be of interest for future studies on the compass model. Our investigation identifies a strong one-dimensional magnetic ordering tendency with a large correlation length as the cause of the unusual scaling and moreover allows for a precise quantification of the anomalous length scale involved.

cond-mat.stat-mech↗
arXiv2009

Finite-Temperature Néel Ordering of Fluctuations in a Plaquette Orbital Model

We present a pseudospin model which should be experimentally accessible using solid-state devices and, being a variation on the compass model, adds to the toolbox for the protection of qubits in the area of quantum information. Using Monte Carlo methods, we find for both classical and quantum spins in two and three dimensions Ising-type Neel ordering of energy fluctuations at finite temperatures without magnetic order. We also readdress the controversy concerning the stability of the ordered state in the presence of quenched impurities and present numerical results which are at clear variance with earlier claims in the literature.

cond-mat.stat-mech↗
arXiv2009

Comprehensive quantum Monte Carlo study of the quantum critical points in planar dimerized/quadrumerized Heisenberg models

We study two planar square lattice Heisenberg models with explicit dimerization or quadrumerization of the couplings in the form of ladder and plaquette arrangements. We investigate the quantum critical points of those models by means of (stochastic series expansion) quantum Monte Carlo simulations as a function of the coupling ratio $α= J^\prime/J$. The critical point of the order-disorder quantum phase transition in the ladder model is determined as $α_\mathrm{c} = 1.9096(2)$ improving on previous studies. For the plaquette model we obtain $α_\mathrm{c} = 1.8230(2)$ establishing a first benchmark for this model from quantum Monte Carlo simulations. Based on those values we give further convincing evidence that the models are in the three-dimensional (3D) classical Heisenberg universality class. The results of this contribution shall be useful as references for future investigations on planar Heisenberg models such as concerning the influence of non-magnetic impurities at the quantum critical point.

cond-mat.str-el↗
arXiv2008

Evidence of Unconventional Universality Class in a Two-Dimensional Dimerized Quantum Heisenberg Model

The two-dimensional $J$-$J^\prime$ dimerized quantum Heisenberg model is studied on the square lattice by means of (stochastic series expansion) quantum Monte Carlo simulations as a function of the coupling ratio \hbox{$α=J^\prime/J$}. The critical point of the order-disorder quantum phase transition in the $J$-$J^\prime$ model is determined as \hbox{$α_\mathrm{c}=2.5196(2)$} by finite-size scaling for up to approximately $10 000$ quantum spins. By comparing six dimerized models we show, contrary to the current belief, that the critical exponents of the $J$-$J^\prime$ model are not in agreement with the three-dimensional classical Heisenberg universality class. This lends support to the notion of nontrivial critical excitations at the quantum critical point.

cond-mat.stat-mech↗
arXiv2008

Monte Carlo simulations of the directional-ordering transition in the two-dimensional classical and quantum compass model

A comprehensive study of the two-dimensional (2D) compass model on the square lattice is performed for classical and quantum spin degrees of freedom using Monte Carlo and quantum Monte Carlo methods. We employ state-of-the-art implementations using Metropolis, stochastic series expansion and parallel tempering techniques to obtain the critical ordering temperatures and critical exponents. In a pre-investigation we reconsider the classical compass model where we study and contrast the finite-size scaling behavior of ordinary periodic boundary conditions against annealed boundary conditions. It is shown that periodic boundary conditions suffer from extreme finite-size effects which might be caused by closed loop excitations on the torus. These excitations also appear to have severe effects on the Binder parameter. On this footing we report on a systematic Monte Carlo study of the quantum compass model. Our numerical results are at odds with recent literature on the subject which we trace back to neglecting the strong finite-size effects on periodic lattices. The critical temperatures are obtained as $T_\mathrm{c}=0.1464(2)J$ and $T_\mathrm{c}=0.055(1)J$ for the classical and quantum version, respectively, and our data support a transition in the 2D Ising universality class for both cases.

cond-mat.stat-mech↗
arXiv2008

Percolation of Vortices in the 3D Abelian Lattice Higgs Model

The compact Abelian Higgs model is simulated on a cubic lattice where it possesses vortex lines and pointlike magnetic monopoles as topological defects. The focus of this high-precision Monte Carlo study is on the vortex network, which is investigated by means of percolation observables. In the region of the phase diagram where the Higgs and confinement phases are separated by a first-order transition, it is shown that the vortices percolate right at the phase boundary, and that the first-order nature of the transition is reflected by the network. In the crossover region, where the phase boundary ceases to be first order, the vortices are shown to still percolate. In contrast to other observables, the percolation observables show finite-size scaling. The exponents characterizing the critical behavior of the vortices in this region are shown to fall in the random percolation universality class.

hep-lat↗
arXiv2005

Kertesz Line in the Three-Dimensional Compact U(1) Lattice Higgs Model

The three-dimensional lattice Higgs model with compact U(1) gauge symmetry and unit charge is investigated by means of Monte Carlo simulations. The full model with fluctuating Higgs amplitude is simulated, and both energy as well as topological observables are measured. The data show a Higgs and a confined phase separated by a well-defined phase boundary, which is argued to be caused by proliferating vortices. For fixed gauge coupling, the phase boundary consists of a line of first-order phase transitions at small Higgs self-coupling, ending at a critical point. The phase boundary then continues as a Kertesz line across which thermodynamic quantities are nonsingular. Symmetry arguments are given to support these findings.

cond-mat.str-el↗