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.

SourceResearch connectionFrom yearThrough year

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

At least 379 records · Page 21

All sources NASA NTRS DOE OSTI

DOE OSTI2022

On-the-fly autonomous control of neutron diffraction via physics-informed Bayesian active learning

We demonstrate the first live, autonomous control over neutron diffraction experiments by developing and deploying ANDiE: the autonomous neutron diffraction explorer. Neutron scattering is a unique and versatile characterization technique for probing the magnetic structure and behavior of materials. However, instruments at neutron scattering facilities in the world is limited, and instruments at such facilities are perennially oversubscribed. We demonstrate a significant reduction in experimental time required for neutron diffraction experiments by implementation of autonomous navigation of measurement parameter space through machine learning. Prior scientific knowledge and Bayesian active learning are used to dynamically steer the sequence of measurements. We show that ANDiE can experimentally determine the magnetic ordering transition of both MnO and Fe 1.09 Te all while providing a fivefold enhancement in measurement efficiency. Furthermore, in a hypothesis testing post-processing step, ANDiE can determine transition behavior from a set of possible physical models. ANDiE's active learning approach is broadly applicable to a variety of neutron-based experiments and can open the door for neutron scattering as a tool of accelerated materials discovery.

36 MATERIALS SCIENCE↗

NASA NTRS1995

Global electric field determination in the Earth's outer magnetosphere using energetic charged particles

Although many properties of the Earth's magnetosphere have been measured and quantified in the past 30 years since it was discovered, one fundamental measurement (for zeroth order MHD equilibrium) has been made infrequently and with poor spatial coverage - the global electric field. This oversight is due in part to the neglect of theorists. However, there is renewed interest in the convection electric field because it is now realized to be central to many magnetospheric processes, including the global MHD equilibrium, reconnection rates, Region 2 Birkeland currents, magnetosphere ionosphere coupling, ring current and radiation belt transport, substorm injections, and several acceleration mechanisms. Unfortunately the standard experimental methods have not been able to synthesize a global field (excepting the pioneering work of McIlwain's geostationary models) and we are left with an overly simplistic theoretical field, the Volland-Stern electric field model. Single point measurements of the plasmapause were used to infer the appropriate amplitudes of this model, parameterized by K(sub p). Although this result was never intended to be the definitive electric field model, it has gone nearly unchanged for 20 years. The analysis of current data sets requires a great deal more accuracy than can be provided by the Volland-Stern model. The variability of electric field shielding has not been properly addressed although effects of penetrating magnetospheric electric fields has been seen in mid-and low-latitude ionospheric data sets. The growing interest in substorm dynamics also requires a much better assessment of the electric fields responsible for particle injections. Thus we proposed and developed algorithms for extracting electric fields from particle data taken in the Earth's magnetosphere. As a test of the effectiveness of these new techniques, we analyzed data taken by the AMPTE/CCE spacecraft in equatorial orbit from 1984 to 1989.

Eastman, Timothy E.↗

NASA NTRS1995

The Effects of Rapid Heating of Soot: Implications When Using Laser-Induced Incandescence for Soot Diagnostics

Recent experimental efforts have exploited the high temporal and spatial resolution of laser-induced incandescence (LII) as both a qualitative and quantitative measure of soot volume fraction. As a relatively new diagnostic technique, issues remain as to appropriate excitation laser intensities and the poential intrusive characteristics of LII. The high temperatures to which the soot is heated may accelerate heterogeneous reactions between the soot and flame gases. Vaporization of soot by high energy pulsed laser light has been theoretically modelled and experimentally observed. Potential physical and/or chemical changes in the laser-heated soot raises the question of how the LII signal depends upon these changes as well as the inferred soot volume fraction. Thus we investigated the effects of high energy pulsed laser light on the soot particles.

Vanderwal, Randy L.↗

DOE OSTI2026

Exploring Resonance Structures in the Partial-Wave Analysis of ¿p0 Photoproduction at GlueX

This thesis studies what happens when a photon (¿) collides with a proton (p) and produces a neutral omega (¿) and pion (p0) pair, with a recoiling proton (p'), expressed as ¿p ¿ ¿p0p'. We study this and other reactions to better understand the strong nuclear force; one of the four fundamental forces that govern all the physics of the universe. This force is specifically responsible for the binding and decay of subatomic particles, such as the ones here. While we understand the ¿ and p0, what we are actually interested in is a short-lived unknown particle X that decays via X ¿ ¿p0. There are a multitude of possible particles X can be, and so our focus in this work is to find out what X is by determining its properties from the particles we measure. We do this via an intricate analysis procedure known as “partial-wave analysis”. By analogy, one can think of our particle detector as a buoy, and the particles we want to analyze (X) as pebbles hitting a pond. The waves created by the pebble will move our buoy, giving us information about the pebble that produced the wave. However, when multiple pebbles hit our pond, the waves overlap and interfere with each other. Our buoy only can measure the complicated interfering result of all the waves. To disentangle this, our partial-wave analysis works by modeling this interference pattern so that we may infer the properties of the pebbles (particles) we produced. In this thesis, we review the relevant experimental history in photoproduction and related production mechanisms, as well as the theoretical foundations that motivate our measurement. We describe the methods used to collect our data at the GlueX experiment stationed at the Jefferson Lab accelerator facility in Newport News, Virginia. We then detail the selections we apply to ensure our events are almost exclusively ¿p ¿ ¿p0p'. We cover the intensity model, how we verify its capabilities, and finally present our results together with systematic studies. Our primary result is the detection of a b1(1235) meson interfering with a wide JPC = 1-- vector state, measured via a mass-independent partial-wave analysis. It provides precise experimental results that can be used as input for theoretical models of the reaction, yielding conclusions about the procedures responsible for how our universe behaves at its most basic level.

Scheuer, Kevin [College of William and Mary, Willi↗

DOE OSTI2020

Modern Monte Carlo methods for efficient uncertainty quantification and propagation: A survey

Uncertainty quantification (UQ) includes the characterization, integration, and propagation of uncertainties that result from stochastic variations and a lack of knowledge or data in the natural world. Monte Carlo (MC) method is a sampling-based approach that has widely used for quantification and propagation of uncertainties. However, the standard MC method is often time-consuming if the simulation-based model is computationally intensive. This article gives an overview of modern MC methods to address the existing challenges of the standard MC in the context of UQ. Specifically, multilevel Monte Carlo (MLMC) extending the concept of control variates achieves a significant reduction of the computational cost by performing most evaluations with low accuracy and corresponding low cost, and relatively few evaluations at high accuracy and corresponding high cost. Multifidelity Monte Carlo (MFMC) accelerates the convergence of standard Monte Carlo by generalizing the control variates with different models having varying fidelities and varying computational costs. Multimodel Monte Carlo method (MMMC), having a different setting of MLMC and MFMC, aims to address the issue of UQ and propagation when data for characterizing probability distributions are limited. Multimodel inference combined with importance sampling is proposed for quantifying and efficiently propagating the uncertainties resulting from small data sets. All of these three modern MC methods achieve a significant improvement of computational efficiency for probabilistic UQ, particularly uncertainty propagation. Finally, an algorithm summary and the corresponding code implementation are provided for each of the modern MC methods. The extension and application of these methods are discussed in detail.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2021

Dispersion calibration for the National Ignition Facility electron–positron–proton spectrometers for intense laser matter interactions

Electron–positron pairs, produced in intense laser–solid interactions, are diagnosed using magnetic spectrometers with image plates, such as the National Ignition Facility Electron–Positron–Proton Spectrometers (EPPSs). Although modeling can help infer the quantitative value, the accuracy of the models needs to be verified to ensure measurement quality. The dispersion of low-energy electrons and positrons may be affected by fringe magnetic fields near the entrance of the EPPS. We have calibrated the EPPS with six electron beams from a Siemens Oncor linear accelerator (linac) ranging in energy from 2.7 MeV to 15.2 MeV as they enter the spectrometer. A Geant4 Tool for Particle Simulation Monte Carlo simulation was set up to match depth dose curves and lateral profiles measured in water at 100 cm source–surface distance. An accurate relationship was established between the bending magnet current setting and the energy of the electron beam at the exit window. The simulations and measurements were used to determine the energy distributions of the six electron beams at the EPPS slit. Furthermore, analysis of the scanned image plates together with the determined energy distribution arriving in the spectrometer provides improved dispersion curves for the EPPS.

47 OTHER INSTRUMENTATION↗

DOE OSTI2025

A convergence metric for counting statistics in time-resolved small angle neutron scattering

Here, this work introduces a model-independent, dimensionless metric for predicting optimal measurement duration in time-resolved small-angle neutron scattering using early-time data. Built on a Gaussian process regression framework, the method reconstructs scattering profiles with quantified uncertainty, even from sparse or noisy measurements. Demonstrated on the EQ-SANS instrument at the Spallation Neutron Source, the approach generalizes to general SANS instruments with a two-dimensional detector. A key result is the discovery of a dimensionless convergence metric revealing a universal power-law scaling in profile evolution across soft matter systems. When time is normalized by a system-specific characteristic time t*, the variation in inferred profiles collapses onto a single curve with an exponent between −2 and −1. This trend emerges within the first ten time steps, enabling early prediction of measurement sufficiency. The method supports real-time experimental optimization and is especially valuable for maximizing efficiency in low-flux environments such as compact accelerator-based neutron sources.

Tung, Chi-Huan [Oak Ridge National Laboratory (ORN↗

NASA NTRS2002

On the Consistency of Gamma-Ray Burst Spectral Indices with the Synchrotron Shock Model

The current scenario for gamma-ray bursts (GRBs) involves internal shocks for the prompt GRB emission phase and external shocks for the afterglow phase. Assuming optically thin synchrotron emission from isotropically distributed energetic shocked electrons, GRB spectra observed with a low-energy power-law spectral index greater than -2/3 (for positive photon number indices E(exp alpha) indicate a problem with this model. For spectra that do not violate this condition, additional tests of the shock model can be made by comparing the low- and high-energy spectral indices, on the basis of the model's assertion that synchrotron emission from a single power-law distribution of electrons is responsible for both the low-energy and the high-energy power-law portions of the spectra. We find in most cases that the inferred relationship between the two spectral indices of observed GRB spectra is inconsistent with the constraints from the simple optically thin synchrotron shock emission model. In this sense, the prompt burst phase is different from the afterglow phase, and this difference may be related to anisotropic distributions of particles or to their continual acceleration in shocks during the prompt phase.

Preece, R. D.↗

DOE OSTI2022

Inferring the Thermal History of the Intergalactic Medium from the Properties of the Hydrogen and Helium Lyα Forest

The filamentary network of intergalactic medium (IGM) gas that gives origin to the Lyα forest in the spectra of distant quasars encodes information on the physics of structure formation and the early thermodynamics of diffuse baryonic material. Here we use a massive suite of more than 400 high-resolution cosmological hydrodynamical simulations run with the Graphics Processing Unit–accelerated code Cholla to study the IGM at high spatial resolution maintained over the entire computational volume. The simulations capture a wide range of possible IGM thermal histories by varying the photoheating and photoionizing background produced by star-forming galaxies and active galactic nuclei. A statistical comparison of synthetic spectra with the observed 1D flux power spectra of hydrogen at redshifts 2.2 ≤ z ≤ 5.0 and with the helium Lyα opacity at redshifts 2.4 < z < 2.9 tightly constrains the photoionization and photoheating history of the IGM. By leveraging the constraining power of the available Lyα forest data to break model degeneracies, we find that the IGM experienced two main reheating events over 1.2 Gyr of cosmic time. For our best-fit model, hydrogen reionization completes by z R ≈ 6.0 with a first IGM temperature peak of T 0 ≃ 1.3 × 10 4 K and is followed by the reionization of He ii that completes by z R ≈ 3.0 and yields a second temperature peak of T 0 ≃ 1.4 × 10 4 K. We discuss how our results can be used to obtain information on the timing and the sources of hydrogen and helium reionization.

79 ASTRONOMY AND ASTROPHYSICS↗

DOE OSTI2025

Block Lanczos algorithm for lattice QCD spectroscopy and matrix elements

Recent work introduced a new framework for analyzing correlation functions with improved convergence and signal-to-noise properties, as well as rigorous quantification of excited-state effects, based on the Lanczos algorithm and spurious eigenvalue filtering with the Cullum-Willoughby test. Here, we extend this framework to the analysis of correlation-function matrices built from multiple interpolating operators in lattice quantum chromodynamics (QCD) by constructing an oblique generalization of the block Lanczos algorithm, as well as a new physically motivated reformulation of the Cullum-Willoughby test that generalizes to block Lanczos straightforwardly. The resulting block Lanczos method directly extends generalized eigenvalue problem (GEVP) methods, which can be viewed as applying a single iteration of block Lanczos. Block Lanczos provides qualitative and quantitative advantages over GEVP methods analogous to the benefits of Lanczos over the standard effective mass, including faster convergence to ground- and excited-state energies, explicitly computable two-sided error bounds, straightforward extraction of matrix elements of external currents, and asymptotically constant signal-to-noise. No fits or statistical inference are required. Proof-of-principle calculations are performed for noiseless mock-data examples as well as two-by-two proton correlation-function matrices in lattice QCD.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

DOE OSTI2025

Simulation-Based Inference for Neutrino Interaction Model Tuning

This project demonstrates, for the first time, the application of simulation-based inference (SBI) techniques to tune neutrino–nucleus interaction models. Using a mock dataset based on the MicroBooNE tuning of the GENIE event generator, our approach employs a Neural Posterior Estimator (NPE) with Masked Autoregressive Flows (MAF) to infer the posterior distributions of key GENIE parameters directly from simulated histograms. The workflow provides a scalable and amortized framework for performing likelihood-free inference in high-dimensional parameter spaces, offering a pathway to more efficient and uncertainty-aware model tuning for next-generation neutrino experiments such as DUNE and SBND.

Tame-Narvaez, KarlaMaria [Fermi National Accelerat↗

NASA NTRS1992

Global electric field determination in the Earth's outer magnetosphere using charged particles

Although many properties of the Earth's magnetosphere have been measured and quantified in the past 30 years since it was discovered, one fundamental (for a zeroeth order magnetohydrodynamic (MHD) equilibrium) measurement was made infrequently and with poor spatial coverage: the global electric field. This oversight is in part due to the difficulty of measuring a plasma electric field, and in part due to the difficulty of measuring a plasma electric field, and in part due to the neglect of theorists. However, there is renewed interest in the convection electric field, since it has been realized that it is vital for understanding many aspects of the magnetosphere: the global MHD equilibrium, reconnection rates, Region 2 Birkeland currents, magnetosphere-ionosphere coupling, ring current and radiation belt transport, substorm injections, acceleration mechanisms, etc. Unfortunately the standard experimental methods have not been able to synthesize a global field (excepting the pioneering work of McIlwain's geostationary models), and we are left with an overly simplistic theoretical field, the Volland-Stern electric field mode. Again, single point measurements of the plasma pause were used to infer the appropriate amplitudes of the model, parameterized by Kp (Maynard & Chen, JGR 1975). Although this result was never intended to be the definitive electric field model, it has gone nearly unchanged for 15 years. However, the data sets being taken today require a great deal more accuracy than can be provided by the Volland-Stern model. Nor has the variability of the electric field shielding been properly addressed, although effects of penetrating magnetospheric electric fields has been seen in mid- and low-latitude ionospheric data sets. The growing interests in substorm dynamics also requires a much better assessment of the electric fields responsible for particle injections. Thus, we proposed and developed algorithms for extracting electric fields from particle data taken in the earth's magnetosphere. As a test of the effectiveness of these techniques, we analyzed data taken by the AMPTE/CCE spacecraft in equatorial orbit between 1984-1988. Some analytic tools had to be developed before construction of computer algorithms, and they are discussed.

Eastman, T.↗

NASA NTRS1984

Physical Properties of Neptune and Triton Inferred from the Orbit of Triton

The orbital motion of Triton was redetermined from photographic data spanning the interval 1899-1981. The resulting ephemeris should provide Triton positions with respect to Neptune with an accuracy of plus or minus 500 km through the end of the century. The following physical results follow from the orbit solution. The inverse mass of Neptune + Triton is solar mass over (c sub N + m sub T) = 19490 plus or minus 40. No acceleration of the mean motion nor orbital eccentricity were detected, thus constraining the tidal dissipation factors of Neptune and Triton to O sub N less than or equal to 650 and QT less than or equal to 10,000. Tidal heating of Triton is presently insignificant. The gravitational harmonic J sub 2 of Neptune is 0.0043 plus or minus 0.0003 if Triton is as massive as m sub T/ M sub N = 0.00128, and if Neptune's spin is prograde with P approximately 18(h), or J sub 2 = 0.0037 plus or minus 0.0002 if Triton is much less massive. Triton undergoes extreme climatic variations due to the combined motion of it's orbit plane and Neptune's orbital motion. Approximately 10% of Triton's surface is presently hidden from diurnal insolation, which may provide a powerful cold trap for atmospheric voltatiles.

A W Harris↗

NASA NTRS1990

The goals of gamma-ray spectroscopy in high energy astrophysics

The use of high resolution gamma-ray spectroscopy in astrophysics is discussed with specific attention given to the application of the Nuclear Astrophysics Explorer (NAE). The gamma-ray lines from nuclear transitions in radionucleic decay and positron annihilation permits the study of current sites, rates and models of nucleosynthesis, and galactic structure. Diffuse galactic emission is discussed, and the high-resolution observations of gamma-ray lines from discrete sites are also described. Interstellar mixing and elemental abundances can also be inferred from high-resolution gamma-ray spectroscopy of nucleosynthetic products. Compact objects can also be examined by means of gamma-ray emissions, allowing better understanding of neutron stars and the accreting black hole near the galactic center. Solar physics can also be investigated by examining such features as solar-flare particle acceleration and atmospheric abundances.

Lingenfelter, Richard E.↗

NASA NTRS2002

The Velocity Distribution of Isolated Radio Pulsars

We infer the velocity distribution of radio pulsars based on large-scale 0.4 GHz pulsar surveys. We do so by modelling evolution of the locations, velocities, spins, and radio luminosities of pulsars; calculating pulsed flux according to a beaming model and random orientation angles of spin and beam; applying selection effects of pulsar surveys; and comparing model distributions of measurable pulsar properties with survey data using a likelihood function. The surveys analyzed have well-defined characteristics and cover approx. 95% of the sky. We maximize the likelihood in a 6-dimensional space of observables P, dot-P, DM, absolute value of b, mu, F (period, period derivative, dispersion measure, Galactic latitude, proper motion, and flux density). The models we test are described by 12 parameters that characterize a population's birth rate, luminosity, shutoff of radio emission, birth locations, and birth velocities. We infer that the radio beam luminosity (i) is comparable to the energy flux of relativistic particles in models for spin-driven magnetospheres, signifying that radio emission losses reach nearly 100% for the oldest pulsars; and (ii) scales approximately as E(exp 1/2) which, in magnetosphere models, is proportional to the voltage drop available for acceleration of particles. We find that a two-component velocity distribution with characteristic velocities of 90 km/ s and 500 km/ s is greatly preferred to any one-component distribution; this preference is largely immune to variations in other population parameters, such as the luminosity or distance scale, or the assumed spin-down law. We explore some consequences of the preferred birth velocity distribution: (1) roughly 50% of pulsars in the solar neighborhood will escape the Galaxy, while approx. 15% have velocities greater than 1000 km/ s (2) observational bias against high velocity pulsars is relatively unimportant for surveys that reach high Galactic absolute value of z distances, but is severe for spatially bounded surveys; (3) an important low-velocity population exists that increases the fraction of neutron stars retained by globular clusters and is consistent with the number of old objects that accrete from the interstellar medium; (4) under standard assumptions for supernova remnant expansion and pulsar spin-down, approx. 10% of pulsars younger than 20 kyr will appear to lie outside of their host remnants. Finally, we comment on the ramifications of our birth velocity distribution for binary survival and the population of inspiraling binary neutron stars relevant to some GRB models and potential sources for LIGO.

Arzoumanian, Z.↗

NASA NTRS2006

"Suzaku Highlight Results on Supernova Remnants"

Highlights of the early Suzaku (formerly Astro-E2) observations of supernova remnants are presented. Suzaku offers unique capabilities for the study of supernova remnants. The unprecedented combination of imaging and spectral resolution below 1 keV in the X-ray Imaging Spectrometer (XIS) makes possible mapping of C, N and O abundances in Galactic remnants of all ages. The first detection of carbon lines in the Cygnus Loop and mapping of the O VII to O VIII ratio in SN 1006 demonstrate this capability. The XIS sensitivity to soft, low surface brightness emission is exemplified by spectroscopy in the 0.3-1.0 keV band of the North Polar Spur and other Galactic ISM structures. Such observations make possible inferences about plasma conditions and abundances. The sensitivity above 6 keV via a combination of the XIS (below 10 keV) and the Hard X-ray Detector (above 10 keV) allows broad band (2-40 keV) spectroscopy and mapping of extended remnants with hard emission components. These components are generally associated with sites of particle acceleration, and measuring their spectral shape potentially provides information about the TeV electron population and its acceleration and energy loss mechanisms. Examples of such remnants observed by Suzaku are the non-thermal emission dominated remnants RX J1713.7-3946 and RX J0852.0-4622, for which flux beyond 30 keV has been detected. The status of the mission and prospects for future groundbreaking observations of supernova remnants will be discussed.

Petre, Robert↗

DOE OSTI2025

Fast and Scalable FFT-Based GPU-Accelerated Algorithms for Block-Triangular Toeplitz Matrices with Application to Linear Inverse Problems Governed by Autonomous Dynamical Systems

In this work, we present an efficient and scalable algorithm for performing matrix-vector multiplications (matvecs) for block Toeplitz matrices. Such matrices, which are shift-invariant with respect to their blocks, arise in the context of solving inverse problems governed by autonomous systems, and time-invariant systems in particular. In this article, we consider inverse problems that infer unknown parameters from observational data of a linear time-invariant dynamical system given in the form of partial differential equations (PDEs). Matrix-free Newton-conjugate-gradient methods are often the gold standard for solving these inverse problems, but they require numerous actions of the Hessian on a vector. Matrix-free adjoint-based Hessian matvecs require solution of a pair of linearized forward/adjoint PDE solves per Hessian action, which may be prohibitive for large-scale inverse problems. Time invariance of the forward PDE problem leads to a block Toeplitz structure of the discretized parameter-to-observable (p2o) map defining the mapping from inputs (parameters) to outputs (observables) of the PDEs. This block Toeplitz structure enables us to exploit two key properties: (1) compact storage of the p2o map and its adjoint, and (2) efficient fast Fourier transform–based Hessian matvecs. The proposed algorithm is mapped onto large multi-GPU clusters and achieves more than 80% of peak bandwidth on NVIDIA A100 GPUs. Excellent weak scaling is shown for up to 48 A100 GPUs. For the targeted problems, the implementation executes Hessian matvecs within fractions of a second, which is orders of magnitude faster than can be achieved by conventional matrix-free Hessian matvecs via forward/adjoint PDE solves.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2025

Even Higher-Level Synthesis: An Exploration of AI Hardware Accelerators using HLS4ML

With the rise of artificial intelligence, the popularization of deep learning, and a constantly evolving industry, the demand for flexible and efficient tools has never been greater. As algorithms grow more complex, their runtime and energy consumption increase exponentially. Customized hardware accelerators, long used for specific mathematical operations, remain essential for managing modern applications' computational and power demands. Hardware accelerators can speed up complex computations by orders of magnitude, but their manual design and verification processes are often challenging and time-consuming. High-Level Synthesis (HLS) provides a solution by transforming high-level algorithm descriptions, typically written in C++ or SystemC, into synthesizable RTL suitable for hardware implementation. This approach reduces development time for RTL engineers while offering flexibility beyond what traditional handwritten RTL can provide. We extended this capability to the machine-learning domain with the open-source framework hls4ml, which allows neural networks trained in Python frameworks like Tensorflow or PyTorch to be synthesized into efficient hardware representations for the traditional FPGA and ASIC flows. This breakthrough addresses the growing need for reduced design turnaround and easy verification of ML hardware accelerators with low latency and power efficiency constraints. During this tutorial, we will demonstrate how Python complements HLS by simplifying the ML design process, bridging the gap between software and hardware development. Attendees will explore how we translate neural networks modeled in Python into fixed-point C++ models suitable for HLS workflows. We will dive into strategies like Value-Range Analysis and Quantization-Aware Training, which optimize these designs for deployment and evaluate their accuracy, power consumption, and energy efficiency. To exemplify these concepts, experts from Fermilab will share their experiences applying this technology to high-energy physics experiments, where real-time, low-latency processing is critical. Over the years, Fermilab engineers have demonstrated how deep neural networks, optimized for hardware using hls4ml, can meet the stringent requirements of trigger systems at the CERN Large Hadron Collider. These systems rely on rapid decision-making to process immense data volumes while retaining only the most relevant events for further analysis. The application of hls4ml has also been extended to innovative technologies like smart pixel arrays. These smart pixels integrate ML inference capabilities directly into sensor devices, enabling localized data processing at the pixel level. This approach drastically reduces the need to transmit raw data to external processing units, significantly decreasing power consumption and latency. By embedding neural networks within the pixel architecture, the smart pixels can identify and prioritize relevant data in real time, providing a highly efficient solution for edge computing in scenarios such as particle detectors and imaging systems. Fermilab's work highlights the potential of hardware-accelerated ML in scenarios where both speed and power efficiency are mission-critical. Through this tutorial, attendees will gain valuable insights into the challenges and solutions of deploying ML in hardware. Understanding how HLS and hls4ml streamline the development of neural network-based hardware accelerators is fundamental for the industry's future. Participants will learn how these technologies are shaping the future of AI and scientific computing.

Di Guglielmo, Giuseppe [Fermilab]↗