Benchmark Studies of Machine Learning Inference using SOFIE

High energy physics (HEP) experiments are facing unprecedented challenges to deal with huge volumes of data. Grid is the excellent technology for it. However, Incompatibility of a variety of grid middewares is an obstacle to exploit the resources on multiple grid infrastructures. This paper firstly introduces HEP experiments and HEP grid, and then reviews the solutions for grid interoperability and interoperations. Because current HEP applications are tightly with gLite middleware, this paper highlights a gateway approach to interoperate with gLite and described it in detail. Direct access to gLite storage system is very important, and a dedicated tool for HEP data sharing is then proposed and implemented. Finally, the implementation of the interoperability of GOS and gLite is introduced.

Comparison of Bayesian and particle swarm algorithms for hyperparameter optimisation in machine learning applications in high energy physics

Machine learning (ML) inference serving systems can schedule requests to improve GPU utilization and to meet service level objectives (SLOs) or deadlines. However, improving GPU utilization may compromise latency-sensitive scheduling, as concurrent tasks contend for GPU resources and thereby introduce interference. Given that interference effects introduce unpredictability in scheduling, neglecting them may compromise SLO or deadline satisfaction. Nevertheless, existing interference prediction approaches remain limited in several respects, which may restrict their usefulness for scheduling. First, they are often coarse-grained, which ignores runtime co-location dynamics and thus restricts their accuracy in interference prediction. Second, they tend to use a static prediction model, which may not effectively cope with different workload characteristics. To this end, we evaluate the potential limitations of existing interference prediction approaches and outline our ongoing work toward achieving efficient ML inference scheduling.

Design study of a low-power, low-noise front-end for multianode silicon drift detectors

Quantum machine learning could possibly become a valuable alternative to classical machine learning for applications in high energy physics by offering computational speedups. In this study, we employ a support vector machine with a quantum kernel estimator (QSVM-Kernel method) to a recent LHC flagship physics analysis: $t\overline{t}H$ (Higgs boson production in association with a top quark pair). In our quantum simulation study using up to 20 qubits and up to $50\phantom{\rule{0.16em}{0ex}}000$ events, the QSVM-Kernel method performs as well as its classical counterparts in three different platforms from Google Tensorflow Quantum, IBM Quantum, and Amazon Braket. Additionally, using 15 qubits and 100 events, the application of the QSVM-Kernel method on the IBM superconducting quantum hardware approaches the performance of a noiseless quantum simulator. Our study confirms that the QSVM-Kernel method can use the large dimensionality of the quantum Hilbert space to replace the classical feature space in realistic physics data sets.

With the development of the new generation of High Energy Physics (HEP) experiments, huge amounts of data are being generated. Efficient parallel algorithms/frameworks and High IO throughput are key to meet the scalability and performance requirements of HEP offline data analysis. Though Hadoop has gained a lot of attention from scientific community for its scalability and parallel computing framework for large data sets, it’s still difficult to make HEP data processing tasks run directly on Hadoop. In this paper we investigate the application of Hadoop to make HEP jobs run on it transparently. Particularly, we discuss a new mechanism to support HEP software to random access data in HDFS. Because HDFS is streaming data stored only supporting sequential write and append. It cannot satisfy HEP jobs to random access data. This new feature allows the Map/Reduce tasks to random read/write on the local file system on data nodes instead of using Hadoop data streaming interface. This makes HEP jobs run on Hadoop possible. We also develop diverse MapReduce model for HEP jobs such as Corsika simulation, ARGO detector simulation and Medea++ reconstruction. And we develop a toolkit for users to submit/query/remove jobs. In addition, we provide cluster monitoring and account system to benefit to the system availability. This work has been in production for HEP experiment to gain about 40,000 CPU hours per month since September, 2016.

Quantum computing, ultra-low-noise sensing, and high-energy physics experiments often rely on superconducting circuits or semiconductor qubits and devices operating at deep cryogenic temperatures (4K and below). Photonic integrated circuits and interconnects have been demonstrated for scalable communications and optical domain transduction in these systems. Due to energy and area constraints, many of these devices need enhanced light-matter interaction, provided by photonic resonators. A key challenge, however, for using these resonators is the sensitivity of resonance wavelength to process variations and thermal fluctuations. While thermo-optical tuning methods are typically employed at room temperature to mitigate this problem, the thermo-optic effect is ineffective at 4K. To address this issue, we demonstrate a non-volatile approach to tune the resonance of photonic resonators using integrated phase-change materials (PCMs) at cryogenic temperatures. In this work, we report a 10 Gb/s free-carrier dispersion based resonant photonic modulator that can be tuned in a non-volatile fashion at sub-4K temperatures using a commercial silicon photonics process. This method paves the way for realizing scalable cryogenic integrated photonics with thousands of resonant devices for quantum and high-energy physics applications.

The development of improved, cost-effective Nb3Sn superconductors would have an immediate benefit for high field magnets in High Energy Physics (HEP) applications. This project will develop and demonstrate an economical powder-in-tube (PIT) Nb3Sn process for use in magnets for future HEP accelerator research. In Phase I, a low-cost, intermetallic, phasepure Cu5Sn4 powder was developed, a PIT Nb3Sn process was designed, and prototype Nb3Sn conductors were fabricated. A cost-performance analysis was conducted, which suggested that the approach will meet DOE goals. In Phase II, powder manufacturing will be optimized, and advanced PIT designs will be fabricated on intermediate size billets. The data will be used to assemble and manufacture a scaled-up prototype conductor. Material will be produced and made available for building and testing prototype cables and test magnets at DOE national laboratories. In addition to HEP applications, an economical Nb3Sn superconductor would have application in high field magnets used in fusion machines and Nuclear Magnetic Resonance (NMR). For fusion machines, a successful demonstration would have enormous economic and social benefits.

Data processing in high energy physics experiments is a multi-tiered process in which raw detector signals are first processed locally into physics objects, and then collated into event records which can be scrutinized by a fast online trigger system. The resulting selection of events are reconstructed and pass through a number of software filters before arriving at a final offline analysis where hard physical constants are extracted. Although sophisticated statistical data analysis techniques are routinely employed high energy physics, the use of statistical signal processing in the field is has until now been rare. Our paper will begin with an overview of a typical high energy physics data acquisition system, outlining the technologies and tradeoffs involved at each stage. We will then proceed to argue that the dominant roles of model dependence and systematic errors in final physics analyses render statistical signal processing techniques largely inapplicable at this level. We observe, however, that at the low-level pattern recognition and event reconstruction levels, statistical signal processing techniques have been making inroads in high energy physics for a number of years, and examples from the literature will be cited. The viability of the technique for second level triggers will be assessed. Parallels to other other approaches, such as neural networks, will also be drawn. It will be argued that the falling cost of computing hardware favors the growth of statistical signal processing methods in high energy physics

A high gain image intensifier system with fast shutter action for application in high energy physics

In High Energy Physics (HEP) experiments, it is useful for physics analysis and outreach if the event display software can provide sophisticated visualization effects. Unity is professional software that provides 3D modeling and animation production. GDML format files are commonly used for detector descriptions in HEP experiments. In this work, we present a method for automating the import of GDML files to the Unity engine modeling through geometric transformation. With the conversion of the file format, the complicated geometric description of the HEP detector in a GDML file can be imported directly into the Unity engine using an FBX file. This format conversion was applied to BESIII experiments to achieve an automated geometry conversion to build geometry models in the Unity engine.

IntroductionReconstructing low-level particle tracks in neutrino physics can address some of the most fundamental questions about the universe. However, processing petabytes of raw data using deep learning techniques poses a challenging problem in the field of High Energy Physics (HEP). In the Exa.TrkX Project, an illustrative HEP application, preprocessed simulation data is fed into a state-of-art Graph Neural Network (GNN) model, accelerated by GPUs. However, limited GPU memory often leads to Out-of-Memory (OOM) exceptions during training, due to the large size of models and datasets. This problem is exacerbated when deploying models on High-Performance Computing (HPC) systems designed for large-scale applications.MethodsWe observe a high workload imbalance issue during GNN model training caused by the irregular sizes of input graph samples in HEP datasets, contributing to OOM exceptions. We aim to scale GNNs on HPC systems, by prioritizing workload balance in graph inputs while maintaining model accuracy. Our paper introduces diverse balancing strategies aimed at decreasing the maximum GPU memory footprint and avoiding the OOM exception, across various datasets.ResultsOur experiments showcase memory reduction of up to 32.14% compared to the baseline. We also demonstrate the proposed strategies can avoid OOM in application. Additionally, we create a distributed multi-GPU implementation using these samplers to demonstrate the scalability of these techniques on the HEP dataset.DiscussionBy assessing the performance of these strategies as data loading samplers across multiple datasets, we can gauge their effectiveness in both single-GPU and distributed environments. Our experiments, conducted on datasets of varying sizes and across multiple GPUs, broaden the applicability of our work to various GNN applications that handle input datasets with irregular graph sizes.

The subject of this special feature in Measurement Science and Technology concerns measurement methods, devices and subsystems, both hardware and software aspects, applied in large experiments of high energy physics (HEP) and superconducting RF accelerator technology (SRF). These experiments concern mainly the physics of elementary particles or the building of new machines and detectors. The papers present practical examples of applied solutions in large, contemporary, international research projects such as HERA, LHC, FLASH, XFEL, ILC and others.These machines are unique in their global scale and consist of extremely dedicated apparatus. The apparatus is characterized by very large dimensions, a considerable use of resources and a high level of overall technical complexity. They possess a large number of measurement channels (ranging from thousands to over 100 million), are characterized by fast of processing of measured data and high measurement accuracies, and work in quite adverse environments. The measurement channels cooperate with a large number of different sensors of momenta, energies, trajectories of elementary particles, electron, proton and photon beam profiles, accelerating fields in resonant cavities, and many others.The provision of high quality measurement systems requires the designers to use only the most up-to-date technical solutions, measurement technologies, components and devices. Research work in these demanding fields is a natural birthplace of new measurement methods, new data processing and acquisition algorithms, complex, networked measurement system diagnostics and monitoring. These developments are taking place in both hardware and software layers. The chief intention of this special feature is that the papers represent equally some of the most current metrology research problems in HEP and SRF. The accepted papers have been divided into four topical groups: superconducting cavities (4 papers), low level RF systems (8 papers), ionizing radiation (5 papers) and HEP experiments (8 papers).The editors would like to thank cordially all the authors who accepted our invitation to present their very recent results. A number of authors of the papers in this issue are active in the 6th European Framework Research Program CARE—Coordinated Accelerators Research in Europe and ELAN—the European Linear Accelerator Network. Some authors are active in research programs of a global extent such as the LHC, ILC and GDE—the Global Design Effort for the International Linear Collider.We also would like to thank personally, as well as on behalf of all the authors, the Editorial Board of Measurement Science and Technology for accepting this very exciting field of contemporary metrology. This field seems to be really a birthplace of a host of new metrological technologies, where the driving force is the incredibly high technical requirements that must soon be fulfilled if we dream of building new accelerators for elementary particles, new biological materials and medicine alike. Special thanks are due to Professor R S Jachowicz of Warsaw University of Technology for initiating this issue and for continuous support and advice during our work.

Custom monolithic electronic circuits have been developed recently for large detector applications in high energy physics where subsystems require tens of thousands of channels of signal processing and data acquisition. In the design and construction of these enormous detectors, it has been found that monolithic circuits offer significant advantages over discrete implementations through increased performance, flexible packaging, lower power and reduced cost per channel. Much of the integrated circuit design for the high energy physics community is directly applicable to intermediate energy heavy-ion and electron physics. This STTR project conducted in collaboration with researchers at the Holifield Radioactive Ion Beam Facility (HRIBF) at Oak Ridge National Laboratory, sought to develop a new integrated circuit chip set for barium fluoride (BaF{sub 2}) detector arrays based upon existing CMOS monolithic circuit designs created for the high energy physics experiments. The work under the STTR Phase 1 demonstrated through the design, simulation, and testing of several prototype chips the feasibility of using custom CMOS integrated circuits for processing signals from BaF{sub 2} detectors. Function blocks including charge-sensitive amplifiers, comparators, one shots, time-to-amplitude converters, analog memory circuits and buffer amplifiers were implemented during Phase 1 effort. Experimental results from bench testing and laboratory testing with sources were documented.

The high-precision measurements of detectors in a High-Energy Physics (HEP) experiment need a continuous sampling of recorded events during collisions. Therefore, significant hardware changes are required to do online data processing due to a large amount of data generated by such detectors. Because of all these changes, a new Detector Control System (DCS) design is required. This paper presents a definition of the software requirements to be considered during the design, integration, and operation of a detector’s DCS into physics data-taking, for continuous and non-continuous measurement conditions in the experiments. For this, the main operating processes, elements, characteristics, and guidelines of the DCS in the most important HEP experiments around the world were analyzed. Additionally, characteristics, functional and non-functional requirements, and use-cases of the main actors involved in the different processes of this control system software are defined. A visual modeling and design tool based on Unified Modeling Language (UML) is used to obtain a description of these requirements.