Search

In this master thesis, we compared the performance of a Neural Network (NN) and a Boosted Decision Tree (BDT) Machine Learning (ML) algorithm for the binary classification of Dark Matter (DM) signal and the Standard Model (SM) background. We conducted searches based on models from different theoretical principles that share common experimental signatures, including three models based on a new $Z'$ vector boson coupled to a DM candidate: the Dark Higgs model, Light Vector model, and an inelastic Effective Field Theory model. We also studied the direct slepton production model in Supersymmetry, and a Two Higgs Doublet Model with an additional pseudoscalar mediator. In the process of networks optimization we explored methods to mitigate the phenomena of missing variables on datasets, as well as how to weigh simulated samples that have negative weights. Our study involved two approaches: a model dependent approach training one BDT for each model and a model independent approach training three BDTs, in kinematically orthogonal regions for all models simultaneously. We demonstrated that the ML model independent approach consistently achieved higher mass exclusion limits for all studied models compared to the model dependent approach. To perform these analyses, proton-proton collision data collected with the ATLAS detector at the Large Hadron Collider during the Run II data taking period (2015-2018), corresponding to an integrated luminosity of 139 fb$^{-1}$, was used. The utilization of the ATLAS data provided a crucial foundation for training and evaluating the ML algorithms, ensuring their relevance and applicability to real-world physics phenomena.

The task of finding the incident particles from the sensor deposits they leave on particle detectors is called event or particle reconstruction. The sensor deposits can be represented generically as a point cloud, with each point corresponding to three spatial dimensions of the sensor location, the energy deposit, and occasionally, also the time of the deposit. As particle detectors become increasingly more complex, ever-more sophisticated methods are needed to perform particle reconstruction. An example is the ongoing High Luminosity (HL) upgrade of the Large Hadron Collider (HL-LHC). The HL-HLC is the most significant milestone in experimental particle physics and aims to deliver an order of magnitude more data rate compared to the current LHC. As part of the upgrade, the endcap calorimeters of the Compact Muon Solenoid (CMS) experiment -- one of the two largest and general-purpose detectors at the LHC -- will be replaced by the radiation-hard High Granularity Calorimeter (HGCAL). The HGCAL will contain $\sim6$ million sensors to achieve the spatial resolution required for reconstructing individual particles in HL-LHC conditions. It has an irregular geometry due to its hexagonal sensors, with sizes varying across the longitudinal and transverse axes. Further, it generates sparse data as less than $10$% of the sensors register positive energy. Reconstruction in this environment, where highly irregular patterns of hits are left by the particles, is an unprecedentedly intractable and compute-intensive pattern recognition problem. This motivates the use of parallelisation-friendly deep learning approaches. More traditional deep learning methods, however, are not feasible for the HGCAL because a regular grid-like structure is assumed in those approaches. In this thesis, a reconstruction algorithm based on a dynamic graph neural network called GravNet is presented. The network is paired with a segmentation technique, Object Condensation, to first perform point-cloud segmentation on the detector hits. The property-prediction capability of the Object Condensation approach is then used for energy regression of the reconstructed particles. A range of experiments are conducted to show that this method works well in conditions expected in the HGCAL i.e., with $200$ simultaneous proton-proton collisions. Parallel algorithms based on Nvidia CUDA are also presented to address the computational challenges of the graph neural network discussed in this thesis. With the optimisations, reconstruction can be performed by this method in approximately $2$ seconds which is suitable considering the computational constraints at the LHC. The presented method is the first-ever example of deep learning based end-to-end calorimetric reconstruction in high occupancy environments. This sets the stage for the next era of particle reconstruction, which is expected to be end-to-end. While this thesis is focused on the HGCAL, the method discussed is general and can be extended not only to other calorimeters but also to other tasks such as track reconstruction.

In the last few years, a technological turnaround has taken place with the so-called Internet of Things (IoT). This has drastically changed the classic view of IT, which until the IoT era had largely consisted of fully-fledged and powerful desktop computers. Compared to classical computers, IoT devices are designed for a specific use. As a result, the hardware and form factor of these devices can be reduced, the price becomes more affordable and, hence, the number of applications increases. Thus, kitchen appliances, thermometers and oscilloscopes become “smarter” and are permanently networked with the Internet. This is taking on greater proportions due to the wide range of possible applications for IoT devices, such as the bring-your-own-device (BYOD) philosophy of many companies, the ever-increasing networking through home automation, the development of smart cities with sensors, and even through the communication of autonomous vehicles with their environment (V2X). IoT devices, however, consist of computer systems on a reduced scale, which - analogous to classical computer systems - are equally vulnerable to computer attacks. However, there is no trivial way to secure IoT devices by installing antivirus programs or anomaly detection. This is aggravated by the fact that IoT manufacturers often focus on implementing profitable functionalities instead of strengthening the security of the devices. In combination with the lack of standards, this leads to an unmanageable diversity of device models, which, for example in large heterogeneous networks, can neither be identified nor critically evaluated in sufficient time without the help of automation. The focus of this dissertation is to create security risk assessments for IoT devices in large heterogeneous networks to estimate their current and future device security risk. For this purpose, we deal with the following research questions: (i) How can device identification be improved to automatically identify different IoT device models in large-scale heterogeneous networks?, (ii) How can gathering of firmware and vulnerability information be automated for known vulnerabilities including patch detection to improve risk assessment results?, (iii) How can knowledge of software vulnerabilities serve to establish risk metrics including a future risk prediction?, and (iv) Do the solutions of research questions 1–3 scale to large heterogeneous networks? We investigate in two exemplary approaches to automate device identification through different information sources. In the following, we show how a data fusion process can be used to increase the device identification rate for a potentially large number of device identification mechanisms. For each device identification, the three most likely device models of being the device under test are further processed by downloading and analyzing their firmware images. This includes vulnerability detection of contained software up to the detection of whether and when manufacturers have fixed vulnerabilities. From the obtained information, the current and future device security risk indicator is calculated per device. In various visualizations, we show these resulting security risk indicators and the evidence for their calculations based on the user’s IT-security experience. The final result is a highly automated process - for technical and non-technical users - which allows a fast, detailed and traceable security assessment for IoT devices. In this work we develop SAFER, a framework for conducting security risk assessments of IoT devices. We evaluate SAFER in the large and heterogeneous network infrastructure of the European Organization for Nuclear Research (CERN), in which about 312,000 network devices are registered. We use SAFER to identify IoT devices in this network and evaluate them in a security critical way. To enable SAFER to provide a comprehensive risk assessment for its users, our framework only needs the host-name of a device to start. In order to evaluate if users can interpret SAFER’s risk assessments in a comprehensible manner, we conducted a study with 10 technical and 10 non-technical CERN employees. The contributions of our work include: (i) the application of a data fusion methodology to device identification mechanisms in order to improve the detection rate, (ii) the automation of the process from firmware acquisition to the automated detection if and when identified software vulnerabilities have been fixed by the device vendor, (iii) the automated calculation of current as well as predicting future risk indicators for IoT devices, and (iv) the evaluation of our SAFER system in its applicability to large heterogeneous networks.

There has been significant work recently in developing machine learning (ML) models in high energy physics (HEP) for tasks such as classification, simulation, and anomaly detection. Often these models are adapted from those designed for datasets in computer vision or natural language processing, which lack inductive biases suited to HEP data, such as equivariance to its inherent symmetries. Such biases have been shown to make models more performant and interpretable, and reduce the amount of training data needed. To that end, we develop the Lorentz group autoencoder (LGAE), an autoencoder model equivariant with respect to the proper, orthochronous Lorentz group $\mathrm{SO}^+(3,1)$, with a latent space living in the representations of the group. We present our architecture and several experimental results on jets at the LHC and find it outperforms graph and convolutional neural network baseline models on several compression, reconstruction, and anomaly detection metrics. We also demonstrate the advantage of such an equivariant model in analyzing the latent space of the autoencoder, which can improve the explainability of potential anomalies discovered by such ML models., 11 pages, 7 figures, 4 tables, and a 3 page appendix

The Earth mover's distance (EMD) is a useful metric for image recognition and classification, but its usual implementations are not differentiable or too slow to be used as a loss function for training other algorithms via gradient descent. In this paper, we train a convolutional neural network (CNN) to learn a differentiable, fast approximation of the EMD and demonstrate that it can be used as a substitute for computing-intensive EMD implementations. We apply this differentiable approximation in the training of an autoencoder-inspired neural network (encoder NN) for data compression at the high-luminosity LHC at CERN. The goal of this encoder NN is to compress the data while preserving the information related to the distribution of energy deposits in particle detectors. We demonstrate that the performance of our encoder NN trained using the differentiable EMD CNN surpasses that of training with loss functions based on mean squared error., 15 pages, 7 figures, submitted to Machine Learning: Science and Technology

The ALICE Inner Tracking System (ITS) team at CERN has completed commissioning the ITS2 detector, related electronic systems, and data-taking for physics has started. With higher rates and Pb-Pb collisions, increased data rates and radiation effects are expected to challenge the system’s stability. The ALICE ITS upgrade project includes a new complex readout system that interfaces several other complex systems, and debugging the readout chain is challenging. This project consists of a requirement specification, technical analysis, and acceptance test from which a command-line utility named fastPASTA has been developed as a final product to verify readout protocols and inspect raw data to debug failures in the readout system. The project is exploratory in assessing the Rust programming language and ecosystem’s impact on developing performant and reusable software. Source code and documentation are available at the GitLab repository [1].

The physics potential of massive liquid argon TPCs in the low-energy regime is still to be fully reaped because few-hits events encode information that can hardly be exploited by conventional classification algorithms. Machine learning (ML) techniques give their best in these types of classification problems. In this paper, we evaluate their performance against conventional (deterministic) algorithms. We demonstrate that both Convolutional Neural Networks (CNN) and Transformer-Encoder methods outperform deterministic algorithms in one of the most challenging classification problems of low-energy physics (single- versus double-beta events). We discuss the advantages and pitfalls of Transformer-Encoder methods versus CNN and employ these methods to optimize the detector parameters, with an emphasis on the DUNE Phase II detectors ("Module of Opportunity").

The high-energy physics community is investigating the feasibility of deploying machine-learning-based solutions on Field-Programmable Gate Arrays (FPGAs) to improve physics sensitivity while meeting data processing latency limitations. In this contribution, we introduce a novel end-to-end procedure that utilizes a machine learning technique called symbolic regression (SR). It searches equation space to discover algebraic relations approximating a dataset. We use PySR (software for uncovering these expressions based on evolutionary algorithm) and extend the functionality of hls4ml (a package for machine learning inference in FPGAs) to support PySR-generated expressions for resource-constrained production environments. Deep learning models often optimise the top metric by pinning the network size because vast hyperparameter space prevents extensive neural architecture search. Conversely, SR selects a set of models on the Pareto front, which allows for optimising the performance-resource tradeoff directly. By embedding symbolic forms, our implementation can dramatically reduce the computational resources needed to perform critical tasks. We validate our procedure on a physics benchmark: multiclass classification of jets produced in simulated proton-proton collisions at the CERN Large Hadron Collider, and show that we approximate a 3-layer neural network with an inference model that has as low as 5 ns execution time (a reduction by a factor of 13) and over 90% approximation accuracy.

Nowadays, we constantly produce data in an unprecedented scale at various domains. In the context of research data, large organizations, like CERN, produce information which is of significant importance and which cannot be reproduced in the future. It is therefore our responsibility to make sure that this information is preserved in a way that it will be available to the future generations. This challenge, which is broadly referred to as digital preservation, has drawn the attention of several researchers and led to the design of a standard for long-term dig- ital data storage, known as the Open Archival Information System (OAIS) standard. Several systems have been developed towards this direction, however these solu- tions were either not fully-compliant with the OAIS standard, some were short term projects that have been decommissioned or they were not open-source and available to the research community. In this context, CERN proposed the Digital Memory project, a digital archiving initiative that should allow researchers to archive their data in a way that it will be accessible in the future. In this thesis, which is part of the Digital Memory project, we confront the aforementioned challenges by proposing an architecture that is fully OAIS-compliant, is integrated with CERN repositories and supports transparency, as the user can easily manage and monitor the actions performed on archival packages. Initially, we implement a tool that can be used to harvest data from various CERN sources like CDS, Indico, CERN Open Data, Gitlab and CodiMD in an OAIS-compliant format called Submission Information Package (SIP). This package can be supplied to the platform in order to create the actual archival packages that can be stored for long term preservation. These packages contain additional metadata and normalization of content that will guarantee long term survival of the information content. Additionally, we show how easy it is for a user to create, monitor and group their archives by using the User Interface. The platform can be easily deployed by anyone on Open- shift with the use of Helm charts. Concerning our evaluation we discuss how the the performance of the platform can be improved and we show that the resulting packages as well as the platform as a whole is fully OAIS-compliant.

This paper outlines the challenge of preserving the vast and diverse amount of digital content generated by the CERN community. First, we define the initial scope of preservation, giving an overview of the digital repositories selected for developing an OAIS-compliant archive at CERN. The OAIS reference model will then be briefly introduced, explaining the fundamental components of a digital preservation system and how they are being implemented. Next, we introduce the CERN Digital Preservation platform, which is designed to orchestrate these components and enable existing information systems to adopt a preservation strategy. We highlight how existing tools and digital preservation solutions, such as Archivematica, are being integrated into this platform.

Cardiovascular disease supposes a substantial fraction of healthcare systems. The invisible nature of these pathologies demands solutions that enable remote monitoring and tracking. Deep Learning (DL) has arisen as a solution in many fields, and in healthcare, multiple successful applications exist for image enhancement and health outside hospitals. However, the computational requirements and the need for large-scale datasets limit DL. Thus, we often offload computation onto server infrastructure, and various Machine-Learning-as-a-Service (MLaaS) platforms emerged from this need. These enable the conduction of heavy computations in a cloud infrastructure, usually equipped with high-performance computing servers. Unfortunately, the technical barriers persist in healthcare ecosystems since sending sensitive data (e.g., medical records or personally identifiable information) to third-party servers involves privacy and security concerns with legal and ethical implications. In the scope of Deep Learning for Healthcare to improve cardiovascular health, Homomorphic Encryption (HE) is a promising tool to enable secure, private, and legal health outside hospitals. Homomorphic Encryption allows for privacy-preserving computations over encrypted data, thus preserving the privacy of the processed information. Efficient HE requires structural optimizations to perform the complex computation of the internal layers. One such optimization is Packed Homomorphic Encryption (PHE), which encodes multiple elements on a single ciphertext, allowing for efficient Single Instruction over Multiple Data (SIMD) operations. However, using PHE in DL circuits is not straightforward, and it demands new algorithms and data encoding, which existing literature has not adequately addressed. To fill this gap, in this work, we elaborate on novel algorithms to adapt the linear algebra operations of DL layers to PHE. Concretely, we focus on Convolutional Neural Networks. We provide detailed descriptions and insights into the different algorithms and efficient inter-layer data format conversion mechanisms. We formally analyze the complexity of the algorithms in terms of performance metrics and provide guidelines and recommendations for adapting architectures that deal with private data. Furthermore, we confirm the theoretical analysis with practical experimentation. Among other conclusions, we prove that our new algorithms speed up the processing of convolutional layers compared to the existing proposals.

There has been a recent explosion in research into machine-learning-based generative modeling to tackle computational challenges for simulations in high energy physics (HEP). In order to use such alternative simulators in practice, we need well-defined metrics to compare different generative models and evaluate their discrepancy from the true distributions. We present the first systematic review and investigation into evaluation metrics and their sensitivity to failure modes of generative models, using the framework of two-sample goodness-of-fit testing, and their relevance and viability for HEP. Inspired by previous work in both physics and computer vision, we propose two new metrics, the Fr\'echet and kernel physics distances (FPD and KPD, respectively), and perform a variety of experiments measuring their performance on simple Gaussian-distributed, and simulated high energy jet datasets. We find FPD, in particular, to be the most sensitive metric to all alternative jet distributions tested and recommend its adoption, along with the KPD and Wasserstein distances between individual feature distributions, for evaluating generative models in HEP. We finally demonstrate the efficacy of these proposed metrics in evaluating and comparing a novel attention-based generative adversarial particle transformer to the state-of-the-art message-passing generative adversarial network jet simulation model. The code for our proposed metrics is provided in the open source JetNet Python library., Comment: 11 pages, 5 figures, 3 tables, and a 5 page appendix

RooFit is a toolkit for statistical modeling and fitting used by most experiments in particle physics. Just as data sets from next-generation experiments grow, processing requirements for physics analysis become more computationally demanding, necessitating performance optimizations for RooFit. One possibility to speed-up minimization and add stability is the use of Automatic Differentiation (AD). Unlike for numerical differentiation, the computation cost scales linearly with the number of parameters, making AD particularly appealing for statistical models with many parameters. In this paper, we report on one possible way to implement AD in RooFit. Our approach is to add a facility to generate C++ code for a full RooFit model automatically. Unlike the original RooFit model, this generated code is free of virtual function calls and other RooFit-specific overhead. In particular, this code is then used to produce the gradient automatically with Clad. Clad is a source transformation AD tool implemented as a plugin to the clang compiler, which automatically generates the derivative code for input C++ functions. We show results demonstrating the improvements observed when applying this code generation strategy to HistFactory and other commonly used RooFit models. HistFactory is the subcomponent of RooFit that implements binned likelihood models with probability densities based on histogram templates. These models frequently have a very large number of free parameters and are thus an interesting first target for AD support in RooFit., 6 pages, 5 figures, 21st International Workshop on Advanced Computing and Analysis Techniques in Physics Research

The performance gap between CPU and memory widens continuously. Choosing the best memory layout for each hardware architecture is increasingly important as more and more programs become memory bound. For portable codes that run across heterogeneous hardware architectures, the choice of the memory layout for data structures is ideally decoupled from the rest of a program. This can be accomplished via a zero-runtime-overhead abstraction layer, underneath which memory layouts can be freely exchanged. We present the Low-Level Abstraction of Memory Access (LLAMA), a C++ library that provides such a data structure abstraction layer with example implementations for multidimensional arrays of nested, structured data. LLAMA provides fully C++ compliant methods for defining and switching custom memory layouts for user-defined data types. The library is extensible with third-party allocators. Providing two close-to-life examples, we show that the LLAMA-generated AoS (Array of Structs) and SoA (Struct of Arrays) layouts produce identical code with the same performance characteristics as manually written data structures. Integrations into the SPEC CPU\textsuperscript{\textregistered} lbm benchmark and the particle-in-cell simulation PIConGPU demonstrate LLAMA's abilities in real-world applications. LLAMA's layout-aware copy routines can significantly speed up transfer and reshuffling of data between layouts compared with naive element-wise copying. LLAMA provides a novel tool for the development of high-performance C++ applications in a heterogeneous environment., Comment: 39 pages, 10 figures, 11 listings

Pushing the boundaries of sciences and providing more advanced services to individuals and communities continuously demand more sophisticated software, specialized hardware, and a growing need for computing power and storage. At the beginning of the 2020s, we are entering a heterogeneous and distributed computing era where resources will be limited and constrained. Grid communities need to adapt their approach: (i) applications need to support various architectures; (ii) workload management systems have to manage various computing paradigms and guarantee the proper execution of the applications, regardless of the constraints of the underlying systems. This thesis focuses on the latter point through the case of the LHCb experiment. The LHCb collaboration currently relies on an infrastructure involving 170 computing centers across the world, the World LHC Computing Grid, to process a growing amount of Monte Carlo simulations, reproducing the experimental conditions of the experiment. Despite its huge size, it will be unable to handle simulations coming from the next LHC runs in a decent time. In the meantime, national science programs are consolidating computing resources and encourage using supercomputers, which provide tremendous computing power but pose higher integration challenges. In this thesis, we propose different approaches to supply distributed and shared computing resources with LHCb tasks. We developed methods to increase the number of computing resource allocations and their duration. It resulted in an improvement of the LHCb job throughput on a grid infrastructure (+40.86\%). We also designed a series of software solutions to address issues in highly-constrained environments that can be found in supercomputers, such as lack of external connectivity and software dependencies. We have applied those concepts to leverage computing power from four partitions of supercomputers ranked in the Top500.

Proposal to study the feasibility to absorb the computing aspect of the MIC service within IT department existing services. Endorsed by the RCS-ICT Steering Committee.

We present a revived version of CERNLIB, the basis for software ecosystems of most of the pre-LHC HEP experiments. The efforts to consolidate CERNLIB are part of the activities of the Data Preservation for High Energy Physics collaboration to preserve data and software of the past HEP experiments. The presented version is based on CERNLIB version 2006 with numerous patches made for compatibility with modern compilers and operating systems. The code is available in the CERN GitLab repository with all the development history starting from the early 1990s. The updates also include a re-implementation of the build system in CMake to ensure CERNLIB compliance with the current best practices and to increase the chances of preserving the code in a compilable state for the decades to come. The revived CERNLIB project also includes updated documentation, which we believe is a cornerstone for any preserved software depending on it., ACAT2022 Proceedings, 5 pages, 2 figures

The estimation of probability density functions is a non trivial task that over the last years has been tackled with machine learning techniques. Successful applications can be obtained using models inspired by the Boltzmann machine (BM) architecture. In this manuscript, the product Jacobi-Theta Boltzmann machine (pJTBM) is introduced as a restricted version of the Riemann-Theta Boltzmann machine (RTBM) with diagonal hidden sector connection matrix. We show that score matching, based on the Fisher divergence, can be used to fit probability densities with the pJTBM more efficiently than with the original RTBM., 7 pages, 3 figures, ACAT22 proceedings

El trabajo de esta tesis se enmarca dentro del proyecto EventIndex del experimento ATLAS, un gran detector de partı́culas del LHC (Gran Colisionador de Hadrones) en el CERN. El objetivo del proyecto es catalogar todas las colisiones de partı́culas, o eventos, registrados en el detector ATLAS y también simulados a lo largo de sus años de funcionamiento. Con este catálogo se pueden caracterizar los datos a nivel de evento para su búsqueda y localización por parte de los usuarios finales. También se pueden realizar comprobaciones en la cadena de registro y reprocesado de los datos, para comprobar su corrección y optimizar futuros procesos. Debido al incremento en las tasas y volumen de datos esperados en el Run 3 (2022-2025) y el HL-LHC (finales de la década del 2020), se requiere un sistema escalable y que simplifique implementaciones anteriores. En esta tesis se presentan las contribuciones al proyecto en las áreas de recolección de datos distribuida, almacenamiento de cantidades masivas de datos y acceso a los mismos. Una pequeña cantidad de información (metadatos) por evento es indexada en el CERN (Tier-0), y de forma distribuida en el grid en todos los centros de computación que forman parte del experimento ATLAS (10 Tier-1, y del orden de 70 Tier-2). En esta tesis se presenta un nuevo modelo de recolección de datos en el grid basado en un object store como almacenamiento temporal, y con selección dinámica de datos para su ingestión en el almacén de datos final. También se presentan las contribuciones a una nueva solución en un único y gran almacén de datos basado en tecnologı́as de macrodatos (Big Data) como HBase/Phoenix, capaz de sostener las tasas y volumen de ingestión de datos requeridos, y que simplifica y soluciona los problemas de las anteriores soluciones hı́bridas. Finalmente, se presenta un marco de computación y herramientas basadas en Spark para el acceso a los datos y la resolución de cargas de trabajo analı́ticas que acceden a grandes cantidades de datos, como el cálculo del solapado (overlaps) entre eventos de distintos datasets, o el cálculo de eventos duplicados. The work presented in this thesis is framed in the context of the EventIndex project of the ATLAS experiment, a big particle detector of the LHC (Large Hadron Collider) at CERN. The objective of the project is to catalog all the particle collisions, or events, recorded at the ATLAS detector and also simulated over the duration of the experiment. With this catalog, data can be characterized at event granularity, important for searching and locating events by the end users. Other automatic checkings can be done in the data reprocessing chain, in order to assure its correctness and optimize future processings. Due to the rise in the production rates and total volume of the data expected for Run 3 (2022-2025) and the HL-LHC (end of the 2020 decade), a scalable system is required also to simplify previous implementations. In this thesis we present the contributions to the project in the areas of distributed data collection, storage of massive volumes of data and access to them. A small quantity of information (metadata) by event is collected from CERN (Tier-0), and distributedly worldwide in the grid in all the computing centers part of the ATLAS Experiment (10 Tier-1, and around 70 Tier-2). We present a new pull model for data collection in the grid with an object store as a temporary store, from where the data can be dynamically retrieved to be ingested at the final backend. We also present the contributions to a big data store using HBase/Phoenix, able to sustain the required data rates and total volume of data, and that simplifies the limitations of the previous hybrid solutions. Finally, we present a computing framework and tools using Spark for the data access, and solving the analytic use cases that access large amounts of data, such as overlaps or duplicate events detection.

More than a thousand 8" silicon sensors will be visually inspected to look for anomalies on their surface during the quality control preceding assembly into the High-Granularity Calorimeter for the CMS experiment at CERN. A deep learning-based algorithm that pre-selects potentially anomalous images of the sensor surface in real time has been developed to automate the visual inspection. The anomaly detection is done by an ensemble of independent deep convolutional neural networks: an autoencoder and a classifier. The performance is evaluated on images acquired in production. The pre-selection reduces the number of images requiring human inspection by 85%, with recall of 97%. Data gathered in production can be used for continuous learning to improve the accuracy incrementally.

This document presents the offline computing resources needed by LHCb in 2024. The computing requests are based on the Computing Model Technical Design Report for the LHCb Upgrade, adjusted to the currently known LHC running schedule and the expected activities to be performed by the LHCb experiment.

This document reports the usage of computing resources by the LHCb collaboration during the period January 1st – December 31st, 2022. The data in the following sections have been compiled from the WLCG Accounting Utility (WAU) portal. For LHCb-specific information, the data is taken from the DIRAC Accounting at the LHCb DIRAC Web portal.

Homomorphic Encryption (HE) is a set of powerful properties of certain cryptosystems that allow for privacy-preserving operation over the encrypted text. Still, HE is not widespread due to limitations in terms of efficiency and usability. Among the challenges of HE, scheme parametrization (i.e., the selection of appropriate parameters within the algorithms) is a relevant multi-faced problem. First, the parametrization needs to comply with a set of properties to guarantee the security of the underlying scheme. Second, parametrization requires a deep understanding of the low-level primitives since the parameters have a confronting impact on the precision, performance, and security of the scheme. Finally, the circuit to be executed influences, and it is influenced by, the parametrization. Thus, there is no general optimal selection of parameters, and this selection depends on the circuit and the scenario of the application. Currently, most of the existing HE frameworks require cryptographers to address these considerations manually. It requires a minimum of expertise acquired through a steep learning curve. In this paper, we propose a unified solution for the aforementioned challenges. Concretely, we present an expert system combining Fuzzy Logic and Linear Programming. The Fuzzy Logic Modules receive a user selection of high-level priorities for the security, efficiency, and performance of the cryptosystem. Based on these preferences, the expert system generates a Linear Programming Model that obtains optimal combinations of parameters by considering those priorities while preserving a minimum level of security for the cryptosystem. We conduct an extended evaluation where we show that an expert system generates optimal parameter selections that maintain user preferences without undergoing the inherent complexity of analyzing the circuit.

Choosing the best memory layout for each hardware architecture is increasingly important as more and more programs become memory bound. For portable codes that run across heterogeneous hardware architectures, the choice of the memory layout for data structures is ideally decoupled from the rest of a program. The low-level abstraction of memory access (LLAMA) is a C++ library that provides a zero-runtime-overhead abstraction layer, underneath which memory mappings can be freely exchanged to customize data layouts, memory access and access instrumentation, focusing on multidimensional arrays of nested, structured data. After its scientific debut, several improvements and extensions have been added to LLAMA. This includes compile-time array extents for zero-memory-overhead views, support for computations during memory access, new mappings for bit-packing, switching types, byte-splitting, memory access instrumentation, and explicit SIMD support. This contribution provides an overview of recent developments in the LLAMA library.

Model compression is instrumental in optimizing deep neural network inference on resource-constrained hardware. The prevailing methods for network compression, namely quantization and pruning, have been shown to enhance efficiency at the cost of performance. Determining the most effective quantization and pruning strategies for individual layers and parameters remains a challenging problem, often requiring computationally expensive and ad hoc numerical optimization techniques. This paper introduces FITCompress, a novel method integrating layer-wise mixed-precision quantization and unstructured pruning using a unified heuristic approach. By leveraging the Fisher Information Metric and path planning through compression space, FITCompress optimally selects a combination of pruning mask and mixed-precision quantization configuration for a given pre-trained model and compression constraint. Experiments on computer vision and natural language processing benchmarks demonstrate that our proposed approach achieves a superior compression-performance trade-off compared to existing state-of-the-art methods. FITCompress stands out for its principled derivation, making it versatile across tasks and network architectures, and represents a step towards achieving optimal compression for neural networks.

Digital preservation is “the combination of policies, strategies and actions that ensure access to digital content over time". It is defined as “the method of keeping digital material alive so that they remain usable as technological advances render original hardware and software specification obsolete.” CERN does not have a formal digital preservation policy or strategy, and actions are limited to bit preservation for data stored at the CERN data center, via disk replication and cold tape storage. Accessibility of digital data relies on information repositories (IRs, e.g. CDS, EDMS, Indico, Zenodo, MyFiles, etc.) where data is generally deposited on IT-managed storage systems. The lack of a digital preservation policy and relevant actions threatens to shorten the lifespan of digital assets produced by the organization. This document defines a strategy for CERN to align with global best practices.

[EN] The Large Hadron Collider (LHC) at CERN has generated a vast amount of information from physics events, reaching peaks of TB of data per day which are then sent to large storage facilities. Traditionally, data processing workflows in the High Energy Physics (HEP) field have leveraged grid computing resources. In this context, users have been responsible for manually parallelising the analysis, sending tasks to computing nodes and aggregating the partial results. Analysis environments in this field have had a common building block in the ROOT software framework. This is the de facto standard tool for storing, processing and visualising HEP data. ROOT offers a modern analysis tool called RDataFrame, which can parallelise computations from a single machine to a distributed cluster while hiding most of the scheduling and result aggregation complexity from users. This is currently done by leveraging Apache Spark as the distributed execution engine, but other alternatives are being explored by HEP research groups. Notably, Dask has rapidly gained popularity thanks to its ability to interface with batch queuing systems, widespread in HEP grid computing facilities. Furthermore, future upgrades of the LHC are expected to bring a dramatic increase in data volumes. This paper presents a novel implementation of the Dask backend for the distributed RDataFrame tool in order to address the aforementioned future trends. The scalability of the tool with both the new backend and the already available Spark backend is demonstrated for the first time on more than two thousand cores, testing a real HEP analysis., Open Access funding provided thanks to the CRUE-CSIC agreement with Springer Nature. This work benefited from the support of grant PID2020-113656RBC22 funded by Ministerio de Ciencia e Innovacion (Spain) MCIN/AEI/10.13039/501100011033.

Much hope for finding new physics phenomena at microscopic scale relies on the observations obtained from High Energy Physics experiments, like the ones performed at the Large Hadron Collider (LHC). However, current experiments do not indicate clear signs of new physics that could guide the development of additional Beyond Standard Model (BSM) theories. Identifying signatures of new physics out of the enormous amount of data produced at the LHC falls into the class of anomaly detection and constitutes one of the greatest computational challenges. In this article, we propose a novel strategy to perform anomaly detection in a supervised learning setting, based on the artificial creation of anomalies through a random process. For the resulting supervised learning problem, we successfully apply classical and quantum Support Vector Classifiers (CSVC and QSVC respectively) to identify the artificial anomalies among the SM events. Even more promising, we find that employing an SVC trained to identify the artificial anomalies, it is possible to identify realistic BSM events with high accuracy. In parallel, we also explore the potential of quantum algorithms for improving the classification accuracy and provide plausible conditions for the best exploitation of this novel computational paradigm., 15 pages, 10 figures

We investigate the ability of an ensemble reservoir computing approach to predict the long-term behaviour of the phase-space region in which the motion of charged particles in hadron storage rings is bounded, the so-called dynamic aperture. Currently, the calculation of the phase-space stability region of hadron storage rings is performed through direct computer simulations, which are resource- and time-intensive processes. Echo State Networks (ESN) are a class of recurrent neural networks that are computationally effective, since they avoid backpropagation and require only cross-validation. Furthermore, they have been proven to be universal approximants of dynamical systems. In this paper, we present the performance reached by ESN based on an ensemble approach for the prediction of the phase-space stability region and compare it with analytical scaling laws based on the stability-time estimate of the Nekhoroshev theorem for Hamiltonian systems. We observe that the proposed ESN approach is capable of effectively predicting the time evolution of the extent of the dynamic aperture, improving the predictions by analytical scaling laws, thus providing an efficient surrogate model. We investigate the ability of an ensemble reservoir computing approach to predict the long-term behaviour of the phase-space region in which the motion of charged particles in hadron storage rings is bounded, the so-called dynamic aperture. Currently, the calculation of the phase-space stability region of hadron storage rings is performed through direct computer simulations, which are resource- and time-intensive processes. Echo State Networks (ESN) are a class of recurrent neural networks that are computationally effective, since they avoid backpropagation and require only cross-validation. Furthermore, they have been proven to be universal approximants of dynamical systems. In this paper, we present the performance reached by ESN based on an ensemble approach for the prediction of the phase-space stability region and compare it with analytical scaling laws based on the stability-time estimate of the Nekhoroshev theorem for Hamiltonian systems. We observe that the proposed ESN approach is capable of effectively predicting the time evolution of the extent of the dynamic aperture, improving the predictions by analytical scaling laws, thus providing an efficient surrogate model.

We here propose a machine learning approach for monitoring particle detectors in real-time. The goal is to assess the compatibility of incoming experimental data with a reference dataset, characterising the data behaviour under normal circumstances, via a likelihood-ratio hypothesis test. The model is based on a modern implementation of kernel methods, nonparametric algorithms that can learn any continuous function given enough data. The resulting approach is efficient and agnostic to the type of anomaly that may be present in the data. Our study demonstrates the effectiveness of this strategy on multivariate data from drift tube chamber muon detectors., Comment: 16 pages, 7 figures

We propose a new strategy for anomaly detection at the LHC based on unsupervised quantum machine learning algorithms. To accommodate the constraints on the problem size dictated by the limitations of current quantum hardware we develop a classical convolutional autoencoder. The designed quantum anomaly detection models, namely an unsupervised kernel machine and two clustering algorithms, are trained to find new-physics events in the latent representation of LHC data produced by the autoencoder. The performance of the quantum algorithms is benchmarked against classical counterparts on different new-physics scenarios and its dependence on the dimensionality of the latent space and the size of the training dataset is studied. For kernel-based anomaly detection, we identify a regime where the quantum model significantly outperforms its classical counterpart. An instance of the kernel machine is implemented on a quantum computer to verify its suitability for available hardware. We demonstrate that the observed consistent performance advantage is related to the inherent quantum properties of the circuit used., {"references":["https://arxiv.org/abs/2301.10780"]}

The increasing luminosities of future data taking at Large Hadron Collider and next generation collider experiments require an unprecedented amount of simulated events to be produced. Such large scale productions demand a significant amount of valuable computing resources. This brings a demand to use new approaches to event generation and simulation of detector responses. In this paper, we discuss the application of generative adversarial networks (GANs) to the simulation of the LHCb experiment events. We emphasize main pitfalls in the application of GANs and study the systematic effects in detail. The presented results are based on the Geant4 simulation of the LHCb Cherenkov detector., Comment: 6 pages, 4 figures

This thesis presents an evaluation of various graphical user interface (GUI) technologies for use in the supervision software of CERN’s Beam Interlock System (BIS). The evaluation is necessary because the current GUItechnology used by the supervision software, JavaFX, is no longer supported by Oracle. It also relies on aninternal library that is no longer maintained. Furthermore, a second version of the BIS is being developed, anda GUI is needed that works with BIS and BIS2 in parallel.Which technology is most suitable for the next version of the BIS supervision GUI?This question was answered by gathering user stories from users and experts. Simple proof of concepts weredeveloped for each technology, in order to identify technical limitations and register the development time. Theuser stories relevant to the choice of technology were listed, and each technology received points depending onwhether they were able to fulfill the requirement or not.Furthermore, in order to evaluate the ease of use, wireframes were created based on some of the user stories.These were qualitatively evaluated with stakeholders, which produced feedback that may be used for furtherdevelopment.Out of the four technologies evaluated, PyQt scored 6, ACW scored 4, WRAP scored 6 and Grafana scored3. WRAP and Grafana were discarded as viable options, since they were not able to fulfill hard requirements.ACW scored lower because it is web based and comes with a bigger maintenance overhead. This thesis was done as a part of a technical studentship at CERN.

We provide details on the implementation of a machine-learning based particle flow algorithm for CMS. The standard particle flow algorithm reconstructs stable particles based on calorimeter clusters and tracks to provide a global event reconstruction that exploits the combined information of multiple detector subsystems, leading to strong improvements for quantities such as jets and missing transverse energy. We have studied a possible evolution of particle flow towards heterogeneous computing platforms such as GPUs using a graph neural network. The machine-learned PF model reconstructs particle candidates based on the full list of tracks and calorimeter clusters in the event. For validation, we determine the physics performance directly in the CMS software framework when the proposed algorithm is interfaced with the offline reconstruction of jets and missing transverse energy. We also report the computational performance of the algorithm, which scales approximately linearly in runtime and memory usage with the input size., 12 pages, 6 figures. Presented at the ACAT 2021: 20th International Workshop on Advanced Computing and Analysis Techniques in Physics Research, Daejeon, Kr, 29 Nov - 3 Dec 2021

Training and Hyperparameter Optimization (HPO) of deep learning-based AI models are often compute resource intensive and calls for the use of large-scale distributed resources as well as scalable and resource efficient hyperparameter search algorithms. This work studies the potential of using model performance prediction to aid the HPO process carried out on High Performance Computing systems. In addition, a quantum annealer is used to train the performance predictor and a method is proposed to overcome some of the problems derived from the current limitations in quantum systems as well as to increase the stability of solutions. This allows for achieving results on a quantum machine comparable to those obtained on a classical machine, showing how quantum computers could be integrated within classical machine learning tuning pipelines. Furthermore, results are presented from the development of a containerized benchmark based on an AI-model for collision event reconstruction that allows us to compare and assess the suitability of different hardware accelerators for training deep neural networks., Comment: 5 pages, 7 figures. Submitted to the proceedings of the ACAT 2022 conference and is to be published in the Journal Of Physics: Conference Series

The new LHCb Vertex Locator (VELO) for LHCb, comprising a new pixel detector and readout electronics, will be installed in 2021 for data taking in Run 3 at the LHC. The electronics centers around the "VeloPix" ASIC at the front-end operating in a trigger-less readout at 40MHz. A custom serializer, called gigabit wireline transmitter (GWT), and associated custom protocol have been designed for the VeloPix. The GWT data are sent from the serializers of the VeloPix at a line rate of 5.12 Gb/s, reaching a total data rate of 2-3 Tb/s for the full VELO detector. Data are sent over 300-m optic-fiber links to the control and readout electronics cards for deserialization and processing in Intel Arria 10 FPGAs. Because of the VeloPix trigger-less design, latency variances up to 12 mu s can occur between adjacent datagrams. It is therefore essential to buffer and synchronize the data in firmware prior to onward propagation or suffer a huge CPU-processing penalty. This article will describe the architecture of the readout firmware in detail with focus given to the resynchronization mechanism and techniques for cauterization. Issues found during readout commissioning, and scaling resource utilization, along with the their solutions, will be illustrated. The latest results of the firmware data-processing chain can be presented as well as the verification procedures employed in simulation. Challenges for the next generation of the detector will also be presented with ideas for a readout processing solution. The new LHCb Vertex Locator (VELO) for LHCb, comprising a new pixel detector and readout electronics, will be installed in 2021 for data taking in Run 3 at the LHC. The electronics centers around the 'VeloPix' ASIC at the front-end operating in a trigger-less readout at 40MHz. A custom serializer, called gigabit wireline transmitter (GWT), and associated custom protocol have been designed for the VeloPix. The GWT data are sent from the serializers of the VeloPix at a line rate of 5.12 Gb/s, reaching a total data rate of 2-3 Tb/s for the full VELO detector. Data are sent over 300-m optic-fiber links to the control and readout electronics cards for deserialization and processing in Intel Arria 10 FPGAs. Because of the VeloPix trigger-less design, latency variances up to 12 $\mu \text{s}$ can occur between adjacent datagrams. It is therefore essential to buffer and synchronize the data in firmware prior to onward propagation or suffer a huge CPU-processing penalty. This article will describe the architecture of the readout firmware in detail with focus given to the resynchronization mechanism and techniques for cauterization. Issues found during readout commissioning, and scaling resource utilization, along with the their solutions, will be illustrated. The latest results of the firmware data-processing chain can be presented as well as the verification procedures employed in simulation. Challenges for the next generation of the detector will also be presented with ideas for a readout processing solution.

The thesis discusses software engineering for heterogeneous control systems based on the OPC UA technology, also known as the IEC-62541 standard. The author researches and proves that software engineering automation for the standard allows to achieve significant economy of manufacturing costs and coherency in distributed multi-processor heterogeneous control systems and maintain them for long life-time. A new software engineering method is proposed which is compliant with the OPC UA and complementary to it in the software creation aspects. A new notation is proposed (including its graphical notation) as well as object-oriented architecture that can be directly applied in actual software projects. The proposed method was used in a software engineering framework called quasar, created by the thesis author with a number of well-documented external contributions. The method and the framework were used to efficiently create many of the key software components of the LHC accelerator experiment control systems at CERN (Geneva, Switzerland). It is also used by commercial companies. Measurements are presented which quantify the reduction of software engineering costs when the method is used. Other non-measurable advantages are presented too.

CERN is the birthplace of the web and a rich source of scientific information, ideas and discoveries. Yet access to the Web and its valuable resources are difficult for many people, with overwhelming evidence of digital barriers in place. This study examines the current status of Web Accessibility at CERN, identifies areas for improvement, discusses factors that may be driving the situation, and finally offers recommendations and guidelines to provide, and promote, an environment where CERN’s science can be available to all, both within CERN’s communities and the general public.

For almost half a century the performance of processors has been improving exponentially, closely following the observations made by Gordon Moore and Robert Dennard. One after the other, both predictions have come to a halt due to the increased complexity in transistor manufacturing, the power and thermal limitations at extremely small-scale technology nodes, and the implications of Amdahl's law to multiprocessing . Nowadays, more than ever, the path to high performance passes through meticulous software optimization, fine tuning, and the design of domain-specific hardware architectures and accelerators. Within the scope of this thesis, we approach High-Performance Computing from two different standpoints. In the first part of the thesis, we bridge the gap between productivity-oriented, high-level programming languages and high-performance computing techniques. The domain of focus is particle accelerator physics, and more specifically beam dynamics. The state-of-art Beam Longitudinal Dynamics simulator BLonD was developed at CERN in 2014, and since then, BLonD has been driving the baseline choices for key-parameters related to the daily operation of the largest circular particle accelerators and their upgrades, as well the research for future machines. We develop a single-node optimized, multi-threaded version of BLonD to accommodate design-space exploration oriented simulation studies. Then, we build a hybrid, MPI-over-OpenMP version to bring the run-time of previously week-long or even month-long simulations down to a few hours. To achieve that, techniques such as intelligent dynamic load-balancing and approximate computing were employed. Finally, to anticipate the demand for ever-growing simulation workloads, we design a distributed, GPU-accelerated version of the code, which delivers more than two orders of magnitude improved latency and throughput compared to the previous state-of-art. All the above technologies and optimizations are developed in a user-friendly way. The dramatic reduction in execution time enables scientists to simulate beam longitudinal dynamics scenarios that combine more complex physics phenomena with finer resolution and larger number of simulated particles. These complex, accurate and fast simulations are essential in the field of beam dynamics to overcome current technological limitations, plan the upcoming upgrades of particle accelerators, and design future machines that will help science advance further. The second part of the thesis is focused on hardware customization to accommodate the needs of modern applications. GPUs, once used for the acceleration of graphic workloads, have now become the dominant platform for general purpose application acceleration. Their processing power and cost-efficiency have led to their adoption in almost every computing domain, including machine learning, scientific computing, and databases. Monitoring the behavior of multiple, GPU-accelerated workloads, the existence of a significant class of kernels was detected, which, due to limited data parallelism fail to support a large degree of Thread-Level Parallelism and hide the latency of memory operations. These kernels seek for more aggressive Instruction-Level Parallelism strategies to improve stall hiding and fill the execution pipeline. This inefficiency is addressed by designing a novel, light-weight Out-Of-Order GPU (LOOG) micro-architecture. LOOG is designed to re-use and re-purpose existing hardware components to minimize the power and area overheads. By exploiting Instruction-Level Parallelism to complement the existing Thread-Level Parallelism execution model, LOOG surpasses both traditional GPU platforms and other prior-art policies. A thorough discussion of LOOG internals and the key design tradeoffs that had to be considered are provided in the thesis. Moreover, an extensive design space exploration is performed to fine tune LOOG and demonstrate its effectiveness when applied on top of a variety of GPU platforms. The LOOG mechanism outperforms conventional platforms by 27.6% and 22.4% in terms of run-time and energy efficiency, respectively. This is a strong indication that LOOG is a promising alternative GPU micro-architecture, which is capable of expanding the applicability of future GPU platforms even further, to new application domains. To summarize, this thesis proposes two approaches to improve the performance in terms of execution time and energy efficiency, anticipating the ever-increasing computing requirements of modern applications. Firstly, we discuss meticulous software customization to take advantage of existing multi-processors and hardware accelerators, while providing an easy-to-use interface to the user-base. Secondly, we explore micro-architectural specializations to adjust to the needs of modern workloads.

The ATLAS EventIndex system comprises the catalogue of all events collected, processed or generated by the ATLAS experiment at the CERN LHC accelerator, and all associated software tools to collect, store and query this information. ATLAS records several billion particle interactions every year of operation, processes them for analysis and generates even larger simulated data samples; a global catalogue is needed to keep track of the location of each event record and be able to search and retrieve specific events for in-depth investigations. Each EventIndex record includes summary information on the event itself and the pointers to the files containing the full event. Most components of the EventIndex system are implemented using BigData open-source tools. This paper describes the architectural choices and their evolution in time, as well as the past, current and foreseen future implementations of all EventIndex components., 21 pages

There is an increase in interest of showing the importance of HPC for Artificial In- telligence and Artificial Intelligence for HPC. HPC centers are now boasting larger and larger compute power, with more centers reaching exascale. CERN is invest- igating the usage of Artificial Intelligence and Machine Learning to augment or replace some of the traditional workflows within the LHC experiments. Advant- ages of Machine Learning and Artificial Intelligence is their highly parallelizable nature on suitable hardware, such as GPUs. MLPF, like every other large scale model, utilizes large compute resources in or- der to become efficient and accurate. Models and their datasets keep increasing in size, further expanding their need for more computer resources. The work in this thesis includes implementing a distributed version of the Graph Neural Network MLPF using the Horovod framework with the aim to scale the applications to exascale-class supercomputers. Horovod is a well-established framework for distributed workloads within the Artificial Intelligence field. Our work uses the Horovod framework to distribute the work across up to 292 su- percomputer nodes each with up to 4 GPUs, i.e. runs with up to over 1100 GPUs on the Jülich Juwels supercomputer. Our work focuses on experiments on the Nvidia Volta and Ampere architecture, as these were the best available to us during this thesis. During these scaling tests we observe that the performance scales well up to 24 nodes. In particular, we see a speedup of up to 20X and on 24 nodes, whereas the speedup reduces to 50X when going to 100 nodes. The thesis also includes comparisons of scaling performances between the Volta and Ampere GPU architectures. Our results also show that the difference between using Volta and Ampere GPUs diminishes as one scales to many nodes, a tend- ency noticed after only 8 nodes a difference becomes only 36 seconds per epoch, whereas the timings on a single node is 485 seconds and 200 seconds, respect- ively for a single Volta and Ampere GPU. Some suggestions for future work are also included.

The High Energy Physics (HEP) research community studies fundamental physics by analyzing petabytes of data produced in particle colliders and taken by particle detectors every year. To process this amount of data, the analysis code developed by scientists is executed on a global network of computing resources. Scientific software maintainers (so-called “software librarians”) bundle a compatible set of needed scientific software into a versioned release. These large and frequently changing stacks are commonly distributed to scientists and worker nodes using the CernVM-File System (CernVM-FS). This distributed file system enables on-demand loading of binaries installed on centrally managed software repositories. The minimal Linux image CernVM provides a portable and reproducible runtime environment for developing and running scientific software. Its key ingredient is the tight coupling with the CernVM-FS client to provide access to the base platform (operating system and tools) as well as the scientific software. To his day, CernVM images are designed to use full virtualization. The goal of this project is to develop a new version of CernVM, CernVM 5, that should deliver all the benefits of the CernVM appliance and should be equally practical as a Linux container and as a full virtual machine. To this end, the CernVM 5 container image consists of a “Just Enough Operating System (JeOS)”, with its contents defined by the HEP_OSlibs meta-package which is commonly used as a base platform in HEP. CernVM 5 further aims to smoothly integrate the CernVM-FS client in various container runtimes. Lastly, CernVM 5 uses special build tools and post-build processing to ensure that scientific software stacks using their custom compilers and build chains can coexist with standard system applications. As a result, CernVM 5 aims to provide a single, minimal container image that can be used as a virtual appliance for mounting the CernVM-FS client and for running and developing HEP application software.

The increasingly growing scale of modern computing infrastructures solicits more ingenious and automatic solutions to their management. Our work focuses on file transfer failures within the Worldwide Large Hadron Collider Computing Grid and proposes a pipeline to support distributed data management operations by suggesting potential issues to investigate. Specifically, we adopt an unsupervised learning approach leveraging Natural Language Processing and Machine Learning tools to automatically parse error messages and group similar failures. The results are presented in the form of a summary table containing the most common textual patterns and time evolution charts. This approach has two main advantages. First, the joint elaboration of the error string and the transfer’s source/destination enables more informative and compact troubleshooting, as opposed to inspecting each site and checking unique messages separately. As a by-product, this also reduces the number of errors to check by some orders of magnitude (from unique error strings to unique categories or patterns). Second, the time evolution plots allow operators to immediately filter out secondary issues (e.g. transient or in resolution) and focus on the most serious problems first (e.g. escalating failures). As a preliminary assessment, we compare our results with the Global Grid User Support ticketing system, showing that most of our suggestions are indeed real issues (direct association), while being able to cover 89% of reported incidents (inverse relationship).

Event generators simulate particle interactions using Monte Carlo techniques, providing the primary connection between experiment and theory in experimental high energy physics. These software packages, which are the first step in the simulation worflow of collider experiments, represent approximately 5 to 20% of the annual WLCG usage for the ATLAS and CMS experiments. With computing architectures becoming more heterogeneous, it is important to ensure that these key software frameworks can be run on future systems, large and small. In this contribution, recent progress on porting and speeding up the Madgraph5_aMC@NLO event generator on hybrid architectures, i.e. CPU with GPU accelerators, is discussed. The main focus of this work has been in the calculation of scattering amplitudes and "matrix elements", which is the computational bottleneck of an event generation application. For physics processes limited to QCD leading order, the code generation toolkit has been expanded to produce matrix element calculations using C++ vector instructions on CPUs and using CUDA for NVidia GPUs, as well as using Alpaka, Kokkos and SYCL for multiple CPU and GPU architectures. Performance is reported in terms of matrix element calculations per time on NVidia, Intel, and AMD devices. The status and outlook for the integration of this work into a production release usable by the LHC experiments, with the same functionalities and very similar user interfaces as the current Fortran version, is also described., 6 pages, 2 figures, 2 tables; submitted to ICHEP2022 proceedings in PoS

In recent years fully-parametric fast simulation methods based on generative models have been proposed for a variety of high-energy physics detectors. By their nature, the quality of data-driven models degrades in the regions of the phase space where the data are sparse. Since machine-learning models are hard to analyse from the physical principles, the commonly used testing procedures are performed in a data-driven way and can't be reliably used in such regions. In our work we propose three methods to estimate the uncertainty of generative models inside and outside of the training phase space region, along with data-driven calibration techniques. A test of the proposed methods on the LHCb RICH fast simulation is also presented., Under review in Journal Of Physics: Conference Series (ACAT-2021)

[EN] Data analysis workflows in High Energy Physics (HEP) read data written in the ROOT columnar format. Such data has traditionally been stored in files that are often read via the network from remote storage facilities, which represents a performance penalty especially for data processing workflows that are I/O bound. To address that issue, this paper presents a new caching mechanism, implemented in the I/O subsystem of ROOT, which is independent of the storage backend used to write the dataset. Notably, it can be used to leverage the speed of high-bandwidth, low-latency object stores. The performance of this caching approach is evaluated by running a real physics analysis on an Intel DAOS cluster, both on a single node and distributed on multiple nodes., This work benefited from the support of the CERN Strategic R&D Programme on Technologies for Future Experiments [1] and from grant PID2020-113656RB-C22 funded by Ministerio de Ciencia e Innovacion MCIN/AEI/10.13039/501100011033. The hardware used to perform the experimental evaluation involving DAOS (HPE Delphi cluster described in Sect. 5.2) was made available thanks to a collaboration agreement with Hewlett-Packard Enterprise (HPE) and Intel. User access to the Virgo cluster at the GSI institute was given for the purpose of running the benchmarks using the Lustre filesystem.

We present a machine learning approach for model-independent new physics searches. The corresponding algorithm is powered by recent large-scale implementations of kernel methods, nonparametric learning algorithms that can approximate any continuous function given enough data. Based on the original proposal by D'Agnolo and Wulzer (arXiv:1806.02350), the model evaluates the compatibility between experimental data and a reference model, by implementing a hypothesis testing procedure based on the likelihood ratio. Model-independence is enforced by avoiding any prior assumption about the presence or shape of new physics components in the measurements. We show that our approach has dramatic advantages compared to neural network implementations in terms of training times and computational resources, while maintaining comparable performances. In particular, we conduct our tests on higher dimensional datasets, a step forward with respect to previous studies., 22 pages, 13 figures

The occurrence of vacuum arcs or radio frequency (rf) breakdowns is one of the most prevalent factors limiting the high-gradient performance of normal conducting rf cavities in particle accelerators. In this paper, we search for the existence of previously unrecognized features related to the incidence of rf breakdowns by applying a machine learning strategy to high-gradient cavity data from CERN’s test stand for the Compact Linear Collider (CLIC). By interpreting the parameters of the learned models with explainable artificial intelligence (AI), we reverse-engineer physical properties for deriving fast, reliable, and simple rule–based models. Based on 6 months of historical data and dedicated experiments, our models show fractions of data with a high influence on the occurrence of breakdowns. Specifically, it is shown that the field emitted current following an initial breakdown is closely related to the probability of another breakdown occurring shortly thereafter. Results also indicate that the cavity pressure should be monitored with increased temporal resolution in future experiments, to further explore the vacuum activity associated with breakdowns. The occurrence of vacuum arcs or radio frequency (rf) breakdowns is one of the most prevalent factors limiting the high-gradient performance of normal conducting rf cavities in particle accelerators. In this paper, we search for the existence of previously unrecognized features related to the incidence of rf breakdowns by applying a machine learning strategy to high-gradient cavity data from CERN's test stand for the Compact Linear Collider (CLIC). By interpreting the parameters of the learned models with explainable artificial intelligence (AI), we reverse-engineer physical properties for deriving fast, reliable, and simple rule-based models. Based on 6 months of historical data and dedicated experiments, our models show fractions of data with a high influence on the occurrence of breakdowns. Specifically, it is shown that the field emitted current following an initial breakdown is closely related to the probability of another breakdown occurring shortly thereafter. Results also indicate that the cavity pressure should be monitored with increased temporal resolution in future experiments, to further explore the vacuum activity associated with breakdowns.

This write-up documents software support for CANopen and ELMB in JCOP experiments with special focus on the next-generation CANopen OPC-UA server.