arXiv subjects

Sascha Caron

Publications and source records attributed to Sascha Caron.

At least 37 records · Page 2Linked to original sources

arXiv2023

Benchmarking energy consumption and latency for neuromorphic computing in condensed matter and particle physics

The massive use of artificial neural networks (ANNs), increasingly popular in many areas of scientific computing, rapidly increases the energy consumption of modern high-performance computing systems. An appealing and possibly more sustainable alternative is provided by novel neuromorphic paradigms, which directly implement ANNs in hardware. However, little is known about the actual benefits of running ANNs on neuromorphic hardware for use cases in scientific computing. Here we present a methodology for measuring the energy cost and compute time for inference tasks with ANNs on conventional hardware. In addition, we have designed an architecture for these tasks and estimate the same metrics based on a state-of-the-art analog in-memory computing (AIMC) platform, one of the key paradigms in neuromorphic computing. Both methodologies are compared for a use case in quantum many-body physics in two dimensional condensed matter systems and for anomaly detection at 40 MHz rates at the Large Hadron Collider in particle physics. We find that AIMC can achieve up to one order of magnitude shorter computation times than conventional hardware, at an energy cost that is up to three orders of magnitude smaller. This suggests great potential for faster and more sustainable scientific computing with neuromorphic hardware.

cs.ET↗

arXiv2022

Machine Learning and LHC Event Generation

First-principle simulations are at the heart of the high-energy physics research program. They link the vast data output of multi-purpose detectors with fundamental theory predictions and interpretation. This review illustrates a wide range of applications of modern machine learning to event generation and simulation-based inference, including conceptional developments driven by the specific requirements of particle physics. New ideas and tools developed at the interface of particle physics and machine learning will improve the speed and precision of forward simulations, handle the complexity of collision data, and enhance inference as an inverse simulation problem.

hep-ph↗

arXiv2022

Non-Standard Neutrino Spectra From Annihilating Neutralino Dark Matter

Neutrino telescope experiments are rapidly becoming more competitive in indirect detection searches for dark matter. Neutrino signals arising from dark matter annihilations are typically assumed to originate from the hadronisation and decay of Standard Model particles. Here we showcase a supersymmetric model, the BLSSMIS, that can simultaneously obey current experimental limits while still providing a potentially observable non-standard neutrino spectrum from dark matter annihilation.

hep-ph↗

arXiv2022

Mixture-of-theories Training: Can We Find New Physics and Anomalies Better by Mixing Physical Theories?

Model-independent search strategies have been increasingly proposed in recent years because on the one hand there has been no clear signal for new physics and on the other hand there is a lack of a highly probable and parameter-free extension of the standard model. For these reasons, there is no simple search target so far. In this work, we try to take a new direction and ask the question: bearing in mind that we have a large number of new physics theories that go beyond the Standard Model and may contain a grain of truth, can we improve our search strategy for unknown signals by using them "in combination"? In particular, we show that a signal hypothesis based on a large, intermingled set of many different theoretical signal models can be a superior approach to find an unknown BSM signal. Applied to a recent data challenge, we show that "mixture-of-theories training" outperforms strategies that optimize signal regions with a single BSM model as well as most unsupervised strategies. Applications of this work include anomaly detection and the definition of signal regions in the search for signals of new physics.

hep-ph↗

arXiv2022

AutoSourceID-Light. Fast Optical Source Localization via U-Net and Laplacian of Gaussian

$\textbf{Aims}$. With the ever-increasing survey speed of optical wide-field telescopes and the importance of discovering transients when they are still young, rapid and reliable source localization is paramount. We present AutoSourceID-Light (ASID-L), an innovative framework that uses computer vision techniques that can naturally deal with large amounts of data and rapidly localize sources in optical images. $\textbf{Methods}$. We show that the AutoSourceID-Light algorithm based on U-shaped networks and enhanced with a Laplacian of Gaussian filter (Chen et al. 1987) enables outstanding performances in the localization of sources. A U-Net (Ronneberger et al. 2015) network discerns the sources in the images from many different artifacts and passes the result to a Laplacian of Gaussian filter that then estimates the exact location. $\textbf{Results}$. Application on optical images of the MeerLICHT telescope demonstrates the great speed and localization power of the method. We compare the results with the widely used SExtractor (Bertin & Arnouts 1996) and show the out-performances of our method. AutoSourceID-Light rapidly detects more sources not only in low and mid crowded fields, but particularly in areas with more than 150 sources per square arcminute.

astro-ph.IM↗

arXiv2021

Rare and Different: Anomaly Scores from a combination of likelihood and out-of-distribution models to detect new physics at the LHC

We propose a new method to define anomaly scores and apply this to particle physics collider events. Anomalies can be either rare, meaning that these events are a minority in the normal dataset, or different, meaning they have values that are not inside the dataset. We quantify these two properties using an ensemble of One-Class Deep Support Vector Data Description models, which quantifies differentness, and an autoregressive flow model, which quantifies rareness. These two parameters are then combined into a single anomaly score using different combination algorithms. We train the models using a dataset containing only simulated collisions from the Standard Model of particle physics and test it using various hypothetical signals in four different channels and a secret dataset where the signals are unknown to us. The anomaly detection method described here has been evaluated in a summary paper [1] where it performed very well compared to a large number of other methods. The method is simple to implement and is applicable to other datasets in other fields as well.

hep-ph↗

arXiv2021

Identification of point sources in gamma rays using U-shaped convolutional neural networks and a data challenge

At GeV energies, the sky is dominated by the interstellar emission from the Galaxy. With limited statistics and spatial resolution, accurately separating point sources is therefore challenging. Here we present the first application of deep learning based algorithms to automatically detect and classify point sources from gamma-ray data. For concreteness we refer to this approach as AutoSourceID. To detect point sources, we utilized U-shaped convolutional networks for image segmentation and {\it k}-means for source clustering and localization. We also explored the Centroid-Net algorithm, which is designed to find and count objects. The training data are based on 9.5 years of exposure from The Fermi Large Area Telescope (Fermi-LAT) and we used source properties of active galactic nuclei (AGNs) and pulsars (PSRs) from the fourth Fermi-LAT source catalog (4FGL) in addition to several models of background interstellar emission. The results of the localization algorithm are fed into a classification neural network that is trained to separate the three general source classes (AGNs, PSRs, and FAKE sources). We compared our localization algorithms qualitatively with traditional methods and find them to have similar detection thresholds. We also demonstrate the robustness of our source localization algorithms to modifications in the interstellar emission models, which presents a clear advantage over traditional methods. The classification network is able to discriminate between the three classes with typical accuracy of $\sim$ 70%, as long as balanced data sets are used in classification training. In https://github.com/bapanes/AutoSourceID, we publish our training data sets and analysis scripts and invite the community to join the data challenge aimed to improve the localization and classification of gamma-ray point sources.

astro-ph.HE↗

arXiv2021

A comparison of optimisation algorithms for high-dimensional particle and astrophysics applications

Optimisation problems are ubiquitous in particle and astrophysics, and involve locating the optimum of a complicated function of many parameters that may be computationally expensive to evaluate. We describe a number of global optimisation algorithms that are not yet widely used in particle astrophysics, benchmark them against random sampling and existing techniques, and perform a detailed comparison of their performance on a range of test functions. These include four analytic test functions of varying dimensionality, and a realistic example derived from a recent global fit of weak-scale supersymmetry. Although the best algorithm to use depends on the function being investigated, we are able to present general conclusions about the relative merits of random sampling, Differential Evolution, Particle Swarm Optimisation, the Covariance Matrix Adaptation Evolution Strategy, Bayesian Optimisation, Grey Wolf Optimisation, and the PyGMO Artificial Bee Colony, Gaussian Particle Filter and Adaptive Memory Programming for Global Optimisation algorithms.

hep-ph↗

arXiv2021

Event Generation and Statistical Sampling for Physics with Deep Generative Models and a Density Information Buffer

We present a study for the generation of events from a physical process with deep generative models. The simulation of physical processes requires not only the production of physical events, but also to ensure these events occur with the correct frequencies. We investigate the feasibility of learning the event generation and the frequency of occurrence with Generative Adversarial Networks (GANs) and Variational Autoencoders (VAEs) to produce events like Monte Carlo generators. We study three processes: a simple two-body decay, the processes $e^+e^-\to Z \to l^+l^-$ and $p p \to t\bar{t} $ including the decay of the top quarks and a simulation of the detector response. We find that the tested GAN architectures and the standard VAE are not able to learn the distributions precisely. By buffering density information of encoded Monte Carlo events given the encoder of a VAE we are able to construct a prior for the sampling of new events from the decoder that yields distributions that are in very good agreement with real Monte Carlo events and are generated several orders of magnitude faster. Applications of this work include generic density estimation and sampling, targeted event generation via a principal component analysis of encoded ground truth data, anomaly detection and more efficient importance sampling, e.g. for the phase space integration of matrix elements in quantum field theories.

hep-ph↗

arXiv2020

Particle spectra from dark matter annihilation: physics modeling and QCD uncertainties

In this talk, we discuss the physics modelling of particle spectra arising from dark matter (DM) annihilation or decay. In the context of the indirect searches of DM, the final state products will, in general, undergo a set of complicated processes such as resonance decays, QED/QCD radiation, hadronisation and hadron decays. This set of processes lead to stable particles (photons, positrons, anti-protons, and neutrinos among others) which travel for very long distances before reaching the detectors. The modelling of their spectra contains some uncertainties which are often neglected in the relevant analyses. We discuss the sources of these uncertainties and estimate their impact on photon energy spectra for benchmark DM scenarios with $m_χ\in [10, 1000]\,$GeV. Instructions for how to retrieve complete tables from Zenodo are also provided.

hep-ph↗

arXiv2020

Interactive web-based visualization of multi-dimensional physical and astronomical data

In this manuscript, we propose to expand the use of scientific repositories such as Zenodo and HEP Data, in particular in order to better examine multi-parametric solutions of physical models. The implementation of interactive web-based visualizations enables fast and comfortable re-analysis and comparisons of phenomenological data. In order to illustrate our point of view, we present some examples and demos for dark matter models, supersymmetry exclusions and LHC simulations.

hep-ex↗

arXiv2020

Combining outlier analysis algorithms to identify new physics at the LHC

The lack of evidence for new physics at the Large Hadron Collider so far has prompted the development of model-independent search techniques. In this study, we compare the anomaly scores of a variety of anomaly detection techniques: an isolation forest, a Gaussian mixture model, a static autoencoder, and a $β$-variational autoencoder (VAE), where we define the reconstruction loss of the latter as a weighted combination of regression and classification terms. We apply these algorithms to the 4-vectors of simulated LHC data, but also investigate the performance when the non-VAE algorithms are applied to the latent space variables created by the VAE. In addition, we assess the performance when the anomaly scores of these algorithms are combined in various ways. Using supersymmetric benchmark points, we find that the logical AND combination of the anomaly scores yielded from algorithms trained in the latent space of the VAE is the most effective discriminator of all methods tested.

hep-ph↗

arXiv2020

Estimating QCD uncertainties in Monte Carlo event generators for gamma-ray dark matter searches

Motivated by the recent galactic center gamma-ray excess identified in the Fermi-LAT data, we perform a detailed study of QCD fragmentation uncertainties in the modeling of the energy spectra of gamma-rays from Dark-Matter (DM) annihilation. When Dark-Matter particles annihilate to coloured final states, either directly or via decays such as $W^{(*)}\to q\bar{q}'$, photons are produced from a complex sequence of shower, hadronisation and hadron decays. In phenomenological studies, their energy spectra are typically computed using Monte Carlo event generators. These results have however intrinsic uncertainties due to the specific model used and the choice of model parameters, which are difficult to asses and which are typically neglected. We derive a new set of hadronisation parameters (tunes) for the \textsc{Pythia~8.2} Monte Carlo generator from a fit to LEP and SLD data at the $Z$ peak. For the first time, we also derive a conservative set of uncertainties on the shower and hadronisation model parameters. Their impact on the gamma-ray energy spectra is evaluated and discussed for a range of DM masses and annihilation channels. The spectra and their uncertainties are also provided in tabulated form for future use. The fragmentation-parameter uncertainties may be useful for collider studies as well.

hep-ph↗

arXiv2020

Deep-learning enhancement of large scale numerical simulations

Traditional simulations on High-Performance Computing (HPC) systems typically involve modeling very large domains and/or very complex equations. HPC systems allow running large models, but limits in performance increase that have become more prominent in the last 5-10 years will likely be experienced. Therefore new approaches are needed to increase application performance. Deep learning appears to be a promising way to achieve this. Recently deep learning has been employed to enhance solving problems that traditionally are solved with large-scale numerical simulations using HPC. This type of application, deep learning for high-performance computing, is the theme of this whitepaper. Our goal is to provide concrete guidelines to scientists and others that would like to explore opportunities for applying deep learning approaches in their own large-scale numerical simulations. These guidelines have been extracted from a number of experiments that have been undertaken in various scientific domains over the last two years, and which are described in more detail in the Appendix. Additionally, we share the most important lessons that we have learned.

cs.CE↗

arXiv2019

Constraining the Parameters of High-Dimensional Models with Active Learning

Constraining the parameters of physical models with $>5-10$ parameters is a widespread problem in fields like particle physics and astronomy. The generation of data to explore this parameter space often requires large amounts of computational resources. The commonly used solution of reducing the number of relevant physical parameters hampers the generality of the results. In this paper we show that this problem can be alleviated by the use of active learning. We illustrate this with examples from high energy physics, a field where simulations are often expensive and parameter spaces are high-dimensional. We show that the active learning techniques query-by-committee and query-by-dropout-committee allow for the identification of model points in interesting regions of high-dimensional parameter spaces (e.g. around decision boundaries). This makes it possible to constrain model parameters more efficiently than is currently done with the most common sampling algorithms and to train better performing machine learning models on the same amount of data. Code implementing the experiments in this paper can be found on GitHub.

cs.LG↗

arXiv2019

The current status of fine-tuning in supersymmetry

In this paper, we minimize and compare two different fine-tuning measures in four high-scale supersymmetric models that are embedded in the MSSM. In addition, we determine the impact of current and future dark matter direct detection and collider experiments on the fine-tuning. We then compare the low-scale electroweak measure with the high-scale Barbieri-Giudice measure, which generally do not agree. However, we find that they do reduce to the same value when the higgsino parameter drives the degree of fine-tuning. Depending on the high-scale model and fine-tuning definition, we find a minimal fine-tuning of $3-38$ (corresponding to $\mathcal{O}(10-1)\%$) for the low-scale measure, and $63-571$ (corresponding to $\mathcal{O}(1-0.1)\%$) for the high-scale measure. In addition, minimally fine-tuned spectra give rise to a dark matter relic density that is between $10^{-3} < Ωh^2 < 1$, when $μ$ determines the minimum of the fine-tuning. We stress that it is too early to conclude on the fate of supersymmetry, based only on the fine-tuning paradigm.

hep-ph↗

arXiv2019

DeepXS: Fast approximation of MSSM electroweak cross sections at NLO

We present a deep learning solution to the prediction of particle production cross sections over a complicated, high-dimensional parameter space. We demonstrate the applicability by providing state-of-the-art predictions for the production of charginos and neutralinos at the Large Hadron Collider (LHC) at the next-to-leading order in the phenomenological MSSM-19 and explicitly demonstrate the performance for $pp\to\tildeχ^+_1\tildeχ^-_1,$ $\tildeχ^0_2\tildeχ^0_2$ and $\tildeχ^0_2\tildeχ^\pm_1$ as a proof of concept which will be extended to all SUSY electroweak pairs. We obtain errors that are lower than the uncertainty from scale and parton distribution functions with mean absolute percentage errors of well below $0.5\,\%$ allowing a safe inference at the next-to-leading order with inference times that improve the Monte Carlo integration procedures that have been available so far by a factor of $\mathcal{O}(10^7)$ from $\mathcal{O}(\rm{min})$ to $\mathcal{O}(μ\rm{s})$ per evaluation.

hep-ph↗

arXiv2018

SPOT: Open Source framework for scientific data repository and interactive visualization

SPOT is an open source and free visual data analytics tool for multi-dimensional data-sets. Its web-based interface allows a quick analysis of complex data interactively. The operations on data such as aggregation and filtering are implemented. The generated charts are responsive and OpenGL supported. It follows FAIR principles to allow reuse and comparison of the published data-sets. The software also support PostgreSQL database for scalability.

cs.HC↗