Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “inference accelerators”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 181 records · Page 10
All sourcesNASA NTRSDOE OSTI
DOE OSTI2024

Integrating ytopt and libEnsemble to autotune OpenMC

Ytopt is a Python machine-learning-based autotuning software package developed within the ECP PROTEAS-TUNE project. The ytopt software adopts an asynchronous search framework that consists of sampling a small number of input parameter configurations and progressively fitting a surrogate model over the input-output space until exhausting the user-defined maximum number of evaluations or the wall-clock time. libEnsemble is a Python toolkit for coordinating workflows of asynchronous and dynamic ensembles of calculations across massively parallel resources developed within the ECP PETSc/TAO project. libEnsemble helps users take advantage of massively parallel resources to solve design, decision, and inference problems and expands the class of problems that can benefit from increased parallelism. In this paper we present our methodology and framework to integrate ytopt and libEnsemble to take advantage of massively parallel resources to accelerate the autotuning process. Specifically, we focus on using the proposed framework to autotune the ECP ExaSMR application OpenMC, an open source Monte Carlo particle transport code. OpenMC has seven tunable parameters some of which have large ranges such as the number of particles in-flight, which is in the range of 100,000 to 8 million, with its default setting of 1 million. Setting the proper combination of these parameter values to achieve the best performance is extremely time-consuming. Therefore, we apply the proposed framework to autotune the MPI/OpenMP offload version of OpenMC based on a user-defined metric such as the figure of merit (FoM) (particles/s) or energy efficiency energy-delay product (EDP) on Crusher at Oak Ridge Leadership Computing Facility. In conclusion, the experimental results show that we achieve the improvement up to 29.49% in FoM and up to 30.44% in EDP.

Autotuning↗
DOE OSTI2021

Role of relativistic laser intensity on isochoric heating of metal wire targets

In a recent experimental campaign, we used laser-accelerated relativistic hot electrons to ensure heating of thin titanium wire targets up to a warm dense matter (WDM) state [EPL 114, 45002 (2016)]. The WDM temperature profiles along several hundred microns of the wire were inferred by using spatially resolved X-ray emission spectroscopy looking at the Ti Kαcharacteristic lines. A maximum temperature of ~30 eV was reached. Our study extends this work by discussing the influence of the laser parameters on temperature profiles and the optimisation of WDM wire-based generation. The depth of wire heating may reach several hundreds of microns and it is proven to be strictly dependent on the laser intensity. At the same time, it is quantitatively demonstrated that the maximum WDM temperature doesn’t appear to be sensitive to the laser intensity and mainly depends on the deposited laser energy considering ranges of 6×10 18 –6×10 20 W/cm 2 and 50–200 J.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2020

Enabling Highly Efficient Capsule Networks Processing Through A PIM-Based Architecture Design

In recent years, the CNNs have achieved great successes in the image processing tasks, e.g., image recognition and object detection. Unfortunately, traditional CNN's classication is found to be easily misled by increasingly complex image features due to the usage of pooling operations, hence unable to preserve accurate position and pose information of the objects. To address this challenge, a novel neural network structure called Capsule Network has been proposed, which introduces equivariance through capsules to signicantly enhance the learning ability for image segmentation and object detection. Due to its requirement of performing a high volume of matrix operations, CapsNets have been generally accelerated on modern GPU platforms that provide highly optimized software library for common deep learning tasks. However, based on our performance characterization on modern GPUs, CapsNets exhibit low effciency due to the special program and execution features of their routing procedure, including massive unshareable intermediate variables and intensive syn- chronizations, which are very dicult to optimize at software level. To address these challenges, we propose a hybrid computing architecture design named PIM-CapsNet. It preserves GPU's on-chip computing capability for accelerating CNN types of layers in CapsNet, while pipelining with an off-chip in-memory acceleration solution that effectively tackles routing procedure's ineffciency by leveraging the processing-in-memory capability of today's 3D stacked memory. Using routing procedure's inherent parallellization feature, our design enables hierarchical improvements on CapsNet inference effciency through minimizing data movement and maximizing parallel processing in memory. Evaluation results demonstrate that our proposed design can achieve substantial improvement on both performance and energy savings for CapsNet inference, with almost zero accuracy loss. The results also suggest good performance scalability in optimizing the routing procedure with increasing network size.

Zhang, Xingyao↗
DOE OSTI2026

Simulation-Based Inference for Neutrino Parameter Tuning

This code trains a simulation-based inference (SBI) model for neutrino interaction parameter tuning and subsequently evaluates its performance on MicroBooNE, T2K, and NuWro datasets.

Tame-Narvaez, Karla [Fermi National Accelerator La↗
DOE OSTI2025

Fermi-LAT Captures a Late-2024 Gamma-Ray Outburst near Two Newly Identified Gamma-Ray-emitting Galactic Supernova Remnants, without Evidence of Physical Association

We report the gamma-ray study of two neighboring in the sky Galactic supernova remnants (SNRs), namely, 3C 434.1 (or G094.0+01.0) and G093.7–00.2. In the vicinity of the two SNRs, a late-2024 gamma-ray outburst—not physically connected to the SNRs—was detected with the Fermi Large Area Telescope (LAT). Employing 16.5 yr of Fermi-LAT data, we report the first firm detection of the gamma-ray counterpart of 3C 434.1 at a 7.2σ significance and a likely detection of an enhanced gamma-ray emission signal at a 4.5σ level from G093.7–00.2. The late-2024 gamma-ray outburst is detected at a ∼12σ significance level over a duration of 3 days. For all three objects, a comprehensive morphological and spectral analysis of the corresponding gamma-ray data was conducted. The prevalent scenario for the nature of the gamma-ray emission of both SNRs is attributed to their surrounding environment. Both are likely interacting with nearby dense material, resulting in hadronically induced gamma-ray emission. We deduce that the outburst source is most likely a blazar in flare, linked to the compact, flat-spectrum radio source TXS 2138+527 (also known as NVSS J213953+530016). Further, the spectral analysis of the gamma-ray source conducted within the 100 MeV–800 GeV energy range using the complete LAT data sets, which yielded a soft averaged spectral shape, supports the hypothesis of a flat spectrum radio quasar (FSRQ) origin for the source; with a most likely leptonically induced scenario for the observed gamma-ray outburst as inferred from the obtained hard spectral shape when utilizing short time intervals focusing on the outburst.

Michailidis, Miltiadis [Stanford Univ., CA (United↗
DOE OSTI2021

No H 0 assistance from assisted quintessence

Early dark energy, as a proposed solution to the Hubble tension, faces an additional “why now” problem. Why should dark energy emerge just prior to recombination, billions of years before the onset of cosmic acceleration? Assisted quintessence explains this connection by positing that multiple scaling fields build up over time to drive the present-day cosmic acceleration. In this framework, early dark energy is inevitable. Yet, we show that scaling also leads to the demise of the scenario: the same feature that solves the coincidence problem then spoils a concordance of the Hubble constant inferred from the cosmic microwave background with that from the local distance ladder. In conclusion, the failure of the model offers a novel lesson on the ability of new physics to resolve the Hubble tension.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2024

Joule-class THz pulses from microchannel targets

Inference of joule-class THz radiation sources from microchannel targets driven with hundreds of joule, picosecond lasers is reported. Here, THz sources of this magnitude are useful for nonlinear pumping of matter and for charged-particle acceleration and manipulation. Microchannel targets demonstrate increased laser–THz conversion efficiency compared to planar foil targets, with laser energy to THz energy conversion up to ~0.9% in the best cases.

47 OTHER INSTRUMENTATION↗
DOE OSTI2022

Deep Learning Based Superconducting Radio-Frequency Cavity Fault Classification at Jefferson Laboratory

This work investigates the efficacy of deep learning (DL) for classifying C100 superconducting radio-frequency (SRF) cavity faults in the Continuous Electron Beam Accelerator Facility (CEBAF) at Jefferson Lab. CEBAF is a large, high-power continuous wave recirculating linac that utilizes 418 SRF cavities to accelerate electrons up to 12 GeV. Recent upgrades to CEBAF include installation of 11 new cryomodules (88 cavities) equipped with a low-level RF system that records RF time-series data from each cavity at the onset of an RF failure. Typically, subject matter experts (SME) analyze this data to determine the fault type and identify the cavity of origin. This information is subsequently utilized to identify failure trends and to implement corrective measures on the offending cavity. Manual inspection of large-scale, time-series data, generated by frequent system failures is tedious and time consuming, and thereby motivates the use of machine learning (ML) to automate the task. This study extends work on a previously developed system based on traditional ML methods (Tennant and Carpenter and Powers and Shabalina Solopova and Vidyaratne and Iftekharuddin, Phys. Rev. Accel. Beams, 2020, 23, 114601), and investigates the effectiveness of deep learning approaches. The transition to a DL model is driven by the goal of developing a system with sufficiently fast inference that it could be used to predict a fault event and take actionable information before the onset (on the order of a few hundred milliseconds). Because features are learned, rather than explicitly computed, DL offers a potential advantage over traditional ML. Specifically, two seminal DL architecture types are explored: deep recurrent neural networks (RNN) and deep convolutional neural networks (CNN). We provide a detailed analysis on the performance of individual models using an RF waveform dataset built from past operational runs of CEBAF. In particular, the performance of RNN models incorporating long short-term memory (LSTM) are analyzed along with the CNN performance. Furthermore, comparing these DL models with a state-of-the-art fault ML model shows that DL architectures obtain similar performance for cavity identification, do not perform quite as well for fault classification, but provide an advantage in inference speed.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2025

Computational tools and data integration to accelerate vaccine development: challenges, opportunities, and future directions

The development of effective vaccines is crucial for combating current and emerging pathogens. Despite significant advances in the field of vaccine development there remain numerous challenges including the lack of standardized data reporting and curation practices, making it difficult to determine correlates of protection from experimental and clinical studies. Significant gaps in data and knowledge integration can hinder vaccine development which relies on a comprehensive understanding of the interplay between pathogens and the host immune system. In this review, we explore the current landscape of vaccine development, highlighting the computational challenges, limitations, and opportunities associated with integrating diverse data types for leveraging artificial intelligence (AI) and machine learning (ML) techniques in vaccine design. We discuss the role of natural language processing, semantic integration, and causal inference in extracting valuable insights from published literature and unstructured data sources, as well as the computational modeling of immune responses. Furthermore, we highlight specific challenges associated with uncertainty quantification in vaccine development and emphasize the importance of establishing standardized data formats and ontologies to facilitate the integration and analysis of heterogeneous data. Through data harmonization and integration, the development of safe and effective vaccines can be accelerated to improve public health outcomes. Looking to the future, we highlight the need for collaborative efforts among researchers, data scientists, and public health experts to realize the full potential of AI-assisted vaccine design and streamline the vaccine development process.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2023

A survey of techniques for optimizing transformer inference

Recent years have seen a phenomenal rise in the performance and applications of transformer neural networks. The family of transformer networks, including Bidirectional Encoder Representations from Transformer (BERT), Generative Pretrained Transformer (GPT) and Vision Transformer (ViT), have shown their effectiveness across Natural Language Processing (NLP) and Computer Vision (CV) domains. Transformer-based networks such as ChatGPT have impacted the lives of common men. However, the quest for high predictive performance has led to an exponential increase in transformers' memory and compute footprint. Researchers have proposed techniques to optimize transformer inference at all levels of abstraction. Further, this paper presents a comprehensive survey of techniques for optimizing the inference phase of transformer networks. We survey techniques such as knowledge distillation, pruning, quantization, neural architecture search and lightweight network design at the algorithmic level. We further review hardware-level optimization techniques and the design of novel hardware accelerators for transformers. We summarize the quantitative results on the number of parameters/FLOPs and the accuracy of several models/techniques to showcase the tradeoff exercised by them. We also outline future directions in this rapidly evolving field of research. We believe that this survey will educate both novice and seasoned researchers and also spark a plethora of research efforts in this field.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2021

Characterization of Self-Magnetic Pinch (SMP) radiographic diode performance on RITS-6 at Sandia National Laboratories: 1) Diode Dynamics, DC Heating to extend Radiation Pulse

Radiographic diodes focus an intense electron beam to a small spot size to minimize the source area of energetic photons for radiographic interrogation. The self-magnetic pinch (SMP) diode has been developed as such a source and operated as a load for the RITS-6 Inductive Voltage Adder (IVA) driver. While experiments support the generally accepted conclusion that a 1:1 aspect diode (cathode diameter equals anode-cathode gap) delivers optimum SMP performance, such experiments also show that reducing the cathode diameter, while reducing spot size, also results in reduced radiation dose, by as much as 50%, and degraded shot reproducibility. Analyzation of the effective electron impingement angle on the anode converter with time made possible by a newly developed dose-rate array diagnostic indicates that fast-developing oscillations of the angle are correlated with early termination of the radiation pulse on many of the smaller-diameter SMP shots. This behavior as a function of relative cathode size persists through experiments with output voltages and currents up to 11.5 MV and 225 kA, respectively, and with spot sizes below ~ few mm. Since simulations to date have not predicted such oscillatory behavior, considerable discussion of the angle-behavior of SMP shots is made to lend credence to the inference. There is clear anecdotal evidence that DC heating of the SMP diode region leads to stabilization of this oscillatory behavior. This is the first of two papers on the performance of the SMP diode on the RITS-6 accelerator.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2021

Real-Time Edge AI for Distributed Systems (READS): Progress on Beam Loss De-Blending for the Fermilab Main Injector and Recycler

The Fermilab Main Injector enclosure houses two accelerators, the Main Injector and Recycler. During normal operation, high intensity proton beams exist simultaneously in both. The two accelerators share the same beam loss monitors (BLM) and monitoring system. Beam losses in the Main Injector enclosure are monitored for tuning the accelerators and machine protection. Losses are currently attributed to a specific machine based on timing. However, this method alone is insufficient and often inaccurate, resulting in more difficult machine tuning and unnecessary machine downtime. Machine experts can often distinguish the correct source of beam loss. This suggests a machine learning (ML) model may be producible to help de-blend losses between machines. Work is underway as part of the Fermilab Real-time Edge AI for Distributed Systems Project (READS) to develop a ML empowered system that collects streamed BLM data and additional machine readings to infer in real-time, which machine generated beam loss.

43 PARTICLE ACCELERATORS↗
DOE OSTI2022

Comprehensive characterization of irradiation induced defects in ceria: Impact of point defects on vibrational and optical properties

Validation of multiscale microstructure evolution models can be improved when standard microstructure characterization tools are coupled with methods sensitive to individual point defects. We demonstrate how electronic and vibrational properties of defects revealed by optical absorption and Raman spectroscopies can be used to compliment transmission electron microscopy (TEM) and x-ray diffraction (XRD) in the characterization of microstructure evolution in ceria under non-equilibrium conditions. Experimental manifestation of non-equilibrium conditions was realized by exposing cerium dioxide (CeO2) to energetic protons at elevated temperature. Two sintered polycrystalline CeO2 samples were bombarded with protons accelerated to a few MeVs. These irradiation conditions produced a microstructure with resolvable extended defects and a significant concentration of point defects. A rate theory (RT) model was parametrized using the results of TEM, XRD, and thermal conductivity measurements to infer point defect concentrations. An abundance of cerium sublattice defects suggested by the RT model is supported by Raman spectroscopy measurements, which show peak shift and broadening of the intrinsic T2g peak and emergence of new defect peaks. Additionally, spectroscopic ellipsometry measurements performed in lieu of optical absorption reveals the presence of Ce3+ ions associated with oxygen vacancies. This work lays the foundation for a coupled approach that considers a multimodal characterization of microstructures to guide and validate complex defect evolution models.

Chauhan, Vinay S.↗
DOE OSTI2025

Measurement of off-shell Higgs boson production in the $H^*\rightarrow ZZ\rightarrow 4\ell$ decay channel using a neural simulation-based inference technique in 13 TeV pp collisions with the ATLAS detector

A measurement of off-shell Higgs boson production in the $H^*\to ZZ\to 4\ell$ decay channel is presented. The measurement uses 140 fb −1 of proton–proton collisions at $\sqrt{s} = 13$ TeV collected by the ATLAS detector at the Large Hadron Collider and supersedes the previous result in this decay channel using the same dataset. The data analysis is performed using a neural simulation-based inference method, which builds per-event likelihood ratios using neural networks. The observed (expected) off-shell Higgs boson production signal strength in the $ZZ\to 4\ell$ decay channel at 68% CL is $0.87^{+0.75}_{-0.54}$ ($1.00^{+1.04}_{-0.95}$ ). The evidence for off-shell Higgs boson production using the $ZZ\to 4\ell$ decay channel has an observed (expected) significance of 2.5σ (1.3σ). The expected result represents a significant improvement relative to that of the previous analysis of the same dataset, which obtained an expected significance of 0.5σ. When combined with the most recent ATLAS measurement in the $ZZ\to 2\ell 2\nu$ decay channel, the evidence for off-shell Higgs boson production has an observed (expected) significance of 3.7σ (2.4σ). The off-shell measurements are combined with the measurement of on-shell Higgs boson production to obtain constraints on the Higgs boson total width. The observed (expected) value of the Higgs boson width at 68% CL is $4.3^{+2.7}_{-1.9}$ ($4.1^{+3.5}_{-3.4}$ ) MeV.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗
DOE OSTI2021

A comparison between residential relocation timing of Sydney and Chicago residents: A Bayesian survival analysis

We report that understanding households' behaviour in residential relocation timing is of great importance in the field of transport engineering and economics. This research aims to develop a residential relocation model by considering the potential dynamic impacts of other households' decisions and variables, including economic and demographic attributes, housing features, intra-household decision-making structures, travel mode choice, and other life-course attributes. A multivariate parametric survival model with both fixed and time-varying covariates is developed. To the best of the authors' knowledge, this study is the first paper in the literature of residential relocation timing to propose the use of a Bayesian model in contrast to the widely used classic frequentist approach and have conducted a discussion on its advantages. An emerging residential relocation dataset collected for two cities in Australia and the USA (Sydney and Chicago cities) has been used, which covers residence, vehicle ownership, occupation, education, economic and demographic attributes of respondents. A comprehensive comparison between the results of two cities and a comparison between two Bayesian and frequentist approaches are made. This study confirms the impact of life-course variables, intra-household decision-making behaviours, and sociodemographic attributes on home mobility. According to the model outputs, the accelerating or decelerating impact of explanatory variables on the relocation timing has been almost the same in the two cities. The Bayesian model was confirmed to have some advantages over the frequentist model, including being straightforward to interpret, availability of making inferences on the results, and ease of handling complex models, and optimisation convergence complexities.

99 GENERAL AND MISCELLANEOUS↗
DOE OSTI2022

Characterization of self-magnetic pinch (SMP) radiographic diode performance on RITS-6 at Sandia National Laboratories. I. Diode dynamics, DC heating to extend radiation pulse

Radiographic diodes focus on an intense electron beam to a small spot size to minimize the source area of energetic photons for radiographic interrogation. The self-magnetic pinch (SMP) diode has been developed as such a source and operated as a load for the six-cavity radiographic integrated test stand (RITS-6) inductive voltage adder driver. While experiments support the generally accepted conclusion that a 1:1 aspect diode (cathode diameter equals anode–cathode gap) delivers optimum SMP performance, such experiments also show that reducing the cathode diameter, while reducing spot size, also results in reduced radiation dose, by as much as 50%, and degraded shot reproducibility. Analysis of the effective electron impingement angle on the anode converter with time made possible by a newly developed dose-rate array diagnostic indicates that fast-developing oscillations of the angle are correlated with early termination of the radiation pulse on many of the smaller-diameter SMP shots. This behavior as a function of relative cathode size persists through experiments with output voltages and currents up to 11.5 MV and 225 kA, respectively, and with spot sizes below approximately few millimeters. Since simulations to date have not predicted such oscillatory behavior, considerable discussion of the angle behavior of SMP shots is made to lend credence to the inference. We report there is clear anecdotal evidence that DC heating of the SMP diode region leads to stabilization of this oscillatory behavior. This is the first of two papers on the performance of the SMP diode on the RITS-6 accelerator.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2022

Semantic Regression for Disentangling Beam Losses in the Fermilab Main Injector and Recycler

Fermilab’s Main Injector enclosure houses two accelerators: the Main Injector (MI) and the Recycler (RR). In periods of joint operation, when both machines contain high intensity beam, radiative beam losses from MI and RR overlap on the enclosure’s beam loss monitoring (BLM) system, making it difficult to attribute those losses to a single machine. Incorrect diagnoses result in unnecessary downtime that incurs both financial and experimental cost. In this work, we introduce a novel neural approach for automatically disentangling each machine’s contributions to those measured losses. Using a continuous adaptation of the popular UNet architecture in conjunction with a novel data augmentation scheme, our model accurately infers the machine of origin on a per-BLM basis in periods of joint and independent operation. Crucially, by extracting beam loss information at varying receptive fields, the method is capable of learning both local and global machine signatures and producing high quality inferences using only raw BLM loss measurements.

43 PARTICLE ACCELERATORS↗
DOE OSTI2026

Spatially resolved polarization swings in the supermassive binary black hole candidate OJ 287 with first Event Horizon Telescope observations

We present the first Event Horizon Telescope 1.3 mm observations of the supermassive binary black hole candidate OJ 287. The observations achieved an unprecedented angular resolution of 18 μas and reveal significant structural and polarization variability over just five days, marking the shortest timescale on which such changes have been directly imaged in this source. The inner jet exhibits a twisted ridgeline structure, with features displaying apparent superluminal motions up to about 22 c. The linear polarization maps reveal three main polarized features whose electric-vector position angles (EVPAs) change substantially over the time span of our observations, including a component with a radial polarization consistent with being produced by a recollimation shock. Most notably, we directly resolved two innermost jet components whose EVPAs rotate in opposite directions. The faster component, moving at 2.4 ± 0.9 μas/day (17.4 ± 6.5 c), exhibits counterclockwise EVPA swings of roughly 3.7° per day, while the slower component, with a proper motion of 1.4 ± 0.3 μas/day (10.2 ± 2.2 c), rotates clockwise at approximately 2.5° per day. Previous studies inferred helical magnetic fields in AGN jets from time-resolved or integrated polarization variability but lacked the angular resolution to directly image this effect. Our results provide spatially resolved evidence that a helical magnetic field threads the jet’s collimation and acceleration zone, ruling out models based on the superposition of unresolved components. Our analysis suggests that propagating shocks interact with a Kelvin–Helmholtz plasma instability, illuminating different phases of the helical magnetic field and producing the observed polarization spatial and temporal variability. Moreover, our model naturally accounts for the more rapid polarization rotation observed in the faster moving component. Our model predicts even more rapid swings in polarization, which could be tested with future observations featuring a more densely sampled time coverage.

OJ 287↗