Search NASA⌕ Search

Engineering topics

Yang, Tingjun

Publications and source records attributed to Yang, Tingjun.

At least 19 records

Accelerating Machine Learning Inference with GPUs in ProtoDUNE Data Processing

Abstract We study the performance of a cloud-based GPU-accelerated inference server to speed up event reconstruction in neutrino data batch jobs. Using detector data from the ProtoDUNE experiment and employing the standard DUNE grid job submission tools, we attempt to reprocess the data by running several thousand concurrent grid jobs, a rate we expect to be typical of current and future neutrino physics experiments. We process most of the dataset with the GPU version of our processing algorithm and the remainder with the CPU version for timing comparisons. We find that a 100-GPU cloud-based server is able to easily meet the processing demand, and that using the GPU version of the event processing algorithm is two times faster than processing these data with the CPU version when comparing to the newest CPUs in our sample. The amount of data transferred to the inference server during the GPU runs can overwhelm even the highest-bandwidth network switches, however, unless care is taken to observe network facility limits or otherwise distribute the jobs to multiple sites. We discuss the lessons learned from this processing campaign and several avenues for future improvements.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

A numerical solver for investigating the space charge effect on the electric field in liquid argon time projection chambers

This paper reports the development of a numerical solver aimed to simulate the interaction between the space charge (i.e. ions) distribution and the electric field in liquid argon time projection chamber (LArTPC) detectors. The ion transport equation is solved by a time-accurate, cell-centered finite volume method and the electric potential equation by a continuous finite element method. The electric potential equation updates the electric field which provides the drift velocity to the ion transport equation. The ion transport equation updates the space charge density distribution which appears as the source term in the electric potential equation. The interaction between the space charge distribution and the electric field is numerically simulated within each physical time step. The convective velocity in the ion transport equation can include the background flow velocity in addition to the electric drift velocity. The numerical solver has been parallelized using the Message Passing Interface (MPI) library. Numerical tests show and verify the capability and accuracy of the current numerical solver. It is planned that the developed numerical solver, together with a Computational Fluid Dynamics (CFD) package which provides the flow velocity field, can be used to investigate the space charge effect on the electric field in large-scale particle detectors.

42 ENGINEERING↗

Measurement of the π - -Ar total hadronic cross section at the LArIAT experiment

We present the first measurement of the negative pion total hadronic cross section on argon in a restricted phase space, which we performed at the Liquid Argon In A Testbeam (LArIAT) experiment. All hadronic reaction channels, as well as hadronic elastic interactions with scattering angle greater than 5° are included. The pions have kinetic energies in the range 100–700 MeV and are produced by a beam of charged particles impinging on a solid target at the Fermilab test beam facility. LArIAT employs a 0.24 ton active mass liquid argon time projection chamber (LArTPC) to measure the pion hadronic interactions. For this measurement, LArIAT has developed the “thin slice method,” a new technique to measure cross sections with LArTPCs. While moderately higher, our measurement of the π - -Ar total hadronic cross section is generally in agreement with the geant4 prediction.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Physics Community Needs, Tools, and Resources for Machine Learning

Machine learning (ML) is becoming an increasingly important component of cutting-edge physics research, but its computational requirements present significant challenges. In this white paper, we discuss the needs of the physics community regarding ML across latency and throughput regimes, the tools and resources that offer the possibility of addressing these needs, and how these can be best utilized and accessed in the coming years.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

Extracting low energy signals from raw LArTPC waveforms using deep learning techniques — A proof of concept

Here, we investigate the feasibility of using deep learning techniques, in the form of a one-dimensional convolutional neural network (1D-CNN), for the extraction of signals from the raw waveforms produced by the individual channels of liquid argon time projection chamber (LArTPC) detectors. A minimal generic LArTPC detector model is developed to generate realistic noise and signal waveforms used to train and test the 1D-CNN, and evaluate its performance on low-level signals. We demonstrate that our approach overcomes the inherent shortcomings of traditional cut-based methods by extending sensitivity to signals with ADC values below their imposed thresholds. This approach exhibits great promise in enhancing the capabilities of future generation neutrino experiments like DUNE to carry out their low-energy neutrino physics programs.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

Extracting low energy signals from raw LArTPC waveforms using deep learning techniques -- A proof of concept

We investigate the feasibility of using deep learning techniques, in the form of a one-dimensional convolutional neural network (1D-CNN), for the extraction of signals from the raw waveforms produced by the individual channels of liquid argon time projection chamber (LArTPC) detectors. A minimal generic LArTPC detector model is developed to generate realistic noise and signal waveforms used to train and test the 1D-CNN, and evaluate its performance on low-level signals. We demonstrate that our approach overcomes the inherent shortcomings of traditional cut-based methods by extending sensitivity to signals with ADC values below their imposed thresholds. This approach exhibits great promise in enhancing the capabilities of future generation neutrino experiments like DUNE to carry out their low-energy neutrino physics programs.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

GPU-Accelerated Machine Learning Inference as a Service for Computing in Neutrino Experiments

Machine learning algorithms are becoming increasingly prevalent and performant in the reconstruction of events in accelerator-based neutrino experiments. These sophisticated algorithms can be computationally expensive. At the same time, the data volumes of such experiments are rapidly increasing. The demand to process billions of neutrino events with many machine learning algorithm inferences creates a computing challenge. We explore a computing model in which heterogeneous computing with GPU coprocessors is made available as a web service. The coprocessors can be efficiently and elastically deployed to provide the right amount of computing for a given processing task. With our approach, Services for Optimized Network Inference on Coprocessors (SONIC), we integrate GPU acceleration specifically for the ProtoDUNE-SP reconstruction chain without disrupting the native computing workflow. With our integrated framework, we accelerate the most time-consuming task, track and particle shower hit identification, by a factor of 17. This results in a factor of 2.7 reduction in the total processing time when compared with CPU-only production. For this particular task, only 1 GPU is required for every 68 CPU threads, providing a cost-effective solution.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Scaling Inference Using Triton to Accelerate Particle Physics at the LHC and DUNE

DUNE and the LHC experiments consist of unique and cutting-edge particle detectors that create massive, complex, and rich datasets with billions of events. They require sophisticated algorithms to reconstruct and interpret the data. Modern machine learning algorithms provide a powerful toolset to detect and classify particles, from familiar image processing convolutional neural networks to newer graph neural network architectures. A full reconstruction of these particle collisions requires novel approaches to handle the computing challenge of processing so much raw data. In a series of studies, physicists from Fermilab, CERN, and university groups explored how to accelerate their data processing using the Triton Inference Server.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

Calibration of calorimetric measurement in a liquid argon time projection chamber

The liquid argon time projection chamber provides high-resolution event images and excellent calorimetric resolution for studying neutrino physics and searching for beyond-standard-model physics. In this article, we review the main physics processes that affect detector response, including the electronics and field responses, space charge effects, electron attachment to impurities, diffusion, and recombination. We describe methods to measure those effects, which are used to calibrate the detector response and convert the measured raw analog-to-digital converter (ADC) counts into the original energy deposition.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

Charge response and energy calibration of ProtoDUNE-SP

ProtoDUNE-SP at the CERN Neutrino Platform is a test bed liquid argon time pro-jection chamber (LArTPC) for the far-detector in the Deep Underground Neutrino Ex-periment (DUNE). Space charge eects, attenuation due to electronegative impurities,diusion, and electronics gain variations cause nonuniformities in charge depositionper unit length (dQ/dx) within a LArTPC. Corrections for space charge eects withmeasured electric-eld maps and attenuation with purity-monitor data were applied.A sample of cosmic-ray muons crossing the cathode is used to calibrate the detectorresponse to get a uniform dQ/dx over space and time. Using cosmic-ray stoppingmuons, which have a well-known energy loss per unit length (dE/dx), the absoluteenergy scale was determined. These calibration factors were uploaded to a databasefor use in further physics analysis.

43 PARTICLE ACCELERATORS↗

Direct Photon Results from CDF

Direct (prompt) photon production is a field of very high interest in hadron colliders. It provides probes to search for new phenomena and to test QCD predictions. In this article, two recent cross-section results for direct photon production using the full CDF Run II data set are presented: diphoton production and photon production in association with a heavy quark.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Photon measurements with the full CDF data set

We present results of the cross section measurement of photon production associated with bottom- and charm-quark production and of the cross section measurement for diphoton production. The measurements involve the full CDF data sample and they are compared with state-of-the-art calculations. The comparisons show only partial agreement between data and calculations.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗