Search NASA⌕ Search

SEARCH · Search NASA

Results for “training efficiency”

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 469 records · Page 26
All sourcesNASA NTRSDOE OSTI
DOE OSTI2024

Efficient First-Order Algorithms for Large-Scale, Non-Smooth Maximum Entropy Models with Application to Wildfire Science

Maximum entropy (MaxEnt) models are a class of statistical models that use the maximum entropy principle to estimate probability distributions from data. Due to the size of modern data sets, MaxEnt models need efficient optimization algorithms to scale well for big data applications. State-of-the-art algorithms for MaxEnt models, however, were not originally designed to handle big data sets; these algorithms either rely on technical devices that may yield unreliable numerical results, scale poorly, or require smoothness assumptions that many practical MaxEnt models lack. In this paper, we present novel optimization algorithms that overcome the shortcomings of state-of-the-art algorithms for training large-scale, non-smooth MaxEnt models. Our proposed first-order algorithms leverage the Kullback–Leibler divergence to train large-scale and non-smooth MaxEnt models efficiently. For MaxEnt models with discrete probability distribution of n elements built from samples, each containing m features, the stepsize parameter estimation and iterations in our algorithms scale on the order of O(mn) operations and can be trivially parallelized. Moreover, the strong ℓ1 convexity of the Kullback–Leibler divergence allows for larger stepsize parameters, thereby speeding up the convergence rate of our algorithms. To illustrate the efficiency of our novel algorithms, we consider the problem of estimating probabilities of fire occurrences as a function of ecological features in the Western US MTBS-Interagency wildfire data set. Our numerical results show that our algorithms outperform the state of the art by one order of magnitude and yield results that agree with physical models of wildfire occurrence and previous statistical analyses of wildfire drivers.

Physics↗
DOE OSTI2024

Supporting Growth of a Skilled Workforce in the Building Energy Efficiency Industry (Final Technical Report)

The Interstate Renewable Energy Council (IREC), having worked for well over a decade to develop national networks and best practices focused on attracting and training high quality workers for the clean energy industries, sought to address shortfalls in skilled workers who build, retrofit, operate, and maintain energy efficient buildings. Through this project, IREC created and promoted a Green Buildings Career Map (GBCM)—a career awareness and recruitment tool—that seeks to boost the nationwide growth of workers entering the energy efficiency (EE) industry. Working with a team of subject matter experts and our partner organizations—Building Performance Association (BPA), Building Performance Institute (BPI), Community Action Partnership (CAP), and the National Institute of Building Sciences (NIBS)—IREC produced an interactive, engaging career map to attract and motivate individuals into pursuing careers in energy efficiency, and in support of the Weatherization Assistance Program. The GBCM was designed for a broad audience including educators, career advisors, employers, policymakers, workforce professionals, and jobseekers who may otherwise have little or no awareness of the exciting and rewarding opportunities within targeted industry sectors. Most individuals seeking careers do not know about the breadth of careers awaiting them in the built environment and the same is true for most career and guidance counselors, educators, and parents. The GBCM provides this broad audience with a vision of the myriad jobs that could lead to a long-standing career in energy efficiency. The GBCM describes diverse occupations across the EE industry, charts possible progression between those occupations, and identifies the training, skills, and credentials necessary to do them well.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗
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]↗
DOE OSTI2025

A physics-constrained deep learning treatment of runaway electron dynamics

An adjoint formulation leveraging a physics-informed neural network (PINN) is employed to advance the density moment of a runaway electron (RE) distribution forward in time. A distinguishing feature of this approach is that once the adjoint problem is solved, its solution can be used to project the RE density forward in time for an arbitrary initial momentum space distribution of REs. Furthermore, by employing a PINN, a parametric solution to the adjoint problem can be learned. Thus, once trained, this adjoint-deep learning framework is able to efficiently project the RE density forward in time across various plasma conditions while still including a fully kinetic description of RE dynamics. As an example application, the temporal evolution of the density of primary electrons is studied, with particular emphasis on evaluating the decay of a RE population when below threshold. Predictions from the adjoint-deep learning framework are found to be in good agreement with a traditional relativistic electron Fokker–Planck solver, for several distinct initial conditions, and across an array of physics parameters. Once trained, the PINN thus provides a means of generating RE density time histories with exceptionally low online execution time.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2025

Self-adaptive weights based on balanced residual decay rate for physics-informed neural networks and deep operator networks

Physics-informed deep learning has emerged as a promising alternative for solving partial differential equations. However, for complex problems, training these networks can still be challenging, often resulting in unsatisfactory accuracy and efficiency. In this work, we demonstrate that the failure of plain physics-informed neural networks arises from the significant discrepancy in the convergence rate of residuals at different training points, where the slowest convergence rate dominates the overall solution convergence. Based on these observations, we propose a pointwise adaptive weighting method that balances the residual decay rate across different training points. The performance of our proposed adaptive weighting method is compared with current state-of-the-art adaptive weighting methods on benchmark problems for both physics-informed neural networks and physics-informed deep operator networks. In conclusion, through extensive numerical results we demonstrate that our proposed approach of balanced residual decay rates offers several advantages, including bounded weights, high prediction accuracy, fast convergence rate, low training uncertainty, low computational cost, and ease of hyperparameter tuning.

Balanced convergence rate↗
DOE OSTI2023

Sparse Convolutional Neural Networks for particle classification in ProtoDUNE-SP events

Deep Learning (DL) methods and Computer Vision are becoming important tools for event reconstruction in particle physics detectors. In this work, we report on the use of submanifold sparse convolutional neural networks (SparseNets) for the classification of track and shower hits from a DUNE prototype liquid-argon detector at CERN (ProtoDUNE-SP). By taking advantage of the three-dimensional nature of the problem we use a set of nine input features to classify sparse and locally dense hits associated to track or shower particles. The SparseNet has been trained on a test sample and shows promising results: efficiencies and purities greater than 90%. This has also been achieved with a considerable speedup and substantially less resource utilization with respect to other DL networks such as graph neural networks. This method offers great scalability advantages for future large neutrino detectors such as the planned DUNE experiment.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗
DOE OSTI2021

Normalizing Flows for Microscopic Many-Body Calculations: An Application to the Nuclear Equation of State

We report that normalizing flows are a class of machine learning models used to construct a complex distribution through a bijective mapping of a simple base distribution. We demonstrate that normalizing flows are particularly well suited as a Monte Carlo integration framework for quantum many-body calculations that require the repeated evaluation of high-dimensional integrals across smoothly varying integrands and integration regions. As an example, we consider the finite-temperature nuclear equation of state. An important advantage of normalizing flows is the ability to build highly expressive models of the target integrand, which we demonstrate enables precise evaluations of the nuclear free energy and its derivatives. Furthermore, we show that a normalizing flow model trained on one target integrand can be used to efficiently calculate related integrals when the temperature, density, or nuclear force is varied. This work will support future efforts to build microscopic equations of state for numerical simulations of supernovae and neutron star mergers that employ state-of-the-art nuclear forces and many-body methods.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗
DOE OSTI2022

Deep Learning-Assisted Real-Time Forward Modeling of Electromagnetic Logging in Complex Formations

Higher dimensional (i.e., 2-D and 3-D) modeling is indispensable to correctly evaluate the responses of electromagnetic logging tools in complex formation environments. However, limited by the high computational cost of rigorous modeling such as finite difference method and the finite element method, the real-time applications in the well logging industry primarily rely on the 1-D forward solver, which would result in erroneous formation evaluation for complex scenarios. As a result, aiming at realizing fast modeling for electromagnetic logging tools in complex formations, this paper proposes a general framework assisted by deep neural networks (DNNs). The framework consists of three modules: earth model classification, parameter extraction, and surrogate construction. Separate DNNs are trained and tested for different modules. The accuracy and efficiency of the DNN assisted fast modeling are validated by several experiments. Here, this study finds that the fast modeling assisted by DNNs is able to calculate the tool responses and reconstruct the subsurface formations in real-time.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2023

Nuclear Criticality Experiments Research Center Futures: A Report of a Workshop held September 6-9, 2022, Los Alamos, NM, November, 2022

Experiments and training with critical assemblies and fissionable material (at or near the critical state) that explore reactivity phenomena are central to a number of national security challenges. From fission energy to nuclear weapons to a broad suite of scientific challenges, it is clear that additional capacity and capability are needed. The National Criticality Experiments Research Center (NCERC) marked 10 years of operations in 2021. This anniversary was an opportune time to celebrate our successes and progress, and to evaluate the remaining and emergent challenges. Against this backdrop, a workshop of approximately 140 national and international leaders in nuclear research was convened in Los Alamos, New Mexico to explore “NCERC Futures.” Therefore, the present workshop focused specifically on needed capabilities and tools to meet the research challenges in eight topical mission areas served by NCERC. As each Topical Group summarized their discussions in the out brief, it was recognized that key challenges could be met through enabling infrastructure investments and new critical assemblies. Enabling infrastructure includes staffing, additional space/buildings, developing an agile bounding safety basis, the ability to keep pace with technological advances in detectors and data acquisition systems (allowing use of those with Bluetooth™ and similar technologies), an expanded set of materials options (especially plutonium), a “Plug and Play” design and implementation mindset, and a facility that enabled free-field measurements. The new critical assemblies that were identified as having the most impact were a bare plutonium (Pu) Critical Assembly, a Horizontal Split Table (HST), a Super Comet, and a Uranium Solution Burst Assembly. NECRC is a unique, one-of-a-kind facility in the United States. If all the improvements were to be made, NCERC would enable the United States and its partners to address many important research questions related to criticality. These include but are not limited to: (1) Covering the entire neutron energy spectrum for both highly enriched uranium (HEU) and Pu in configurations for virtually all conceivable applications; (2) Performing multi-physics solution experiments and irradiations with a Uranium Solution Burst Assembly, which more closely resembles actual criticality accidents; (3) Conducting free-field experiments to make basic fission physics measurements and much cleaner benchmarks with various experimental observables; and (4) Providing more training classes and more experiments annually at greater cost efficiency enabled by additional buildings and machines and an agile, bounding, risk-balanced Safety Basis. In the end, workshop attendees enthusiastically concluded that NCERC Futures are bright and the workshop helped to identify a roadmap of capability gaps that need to be addressed. This report documents the results of those efforts.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗
DOE OSTI2023

Clean Energy Employment Impacts and Occupational Analyses: Building Envelope & Electrification Upgrades

Under the Communities LEAP (Local Energy Action Program) Pilot, the U.S. Department of Energy (DOE) is working with a group of LEAP communities to bolster workforce development initiatives and support the growth of local employment in the clean energy economy. Through this resource, DOE seeks to provide participating communities with an overview of the types of occupations that could be affected by investments in building envelope and building electrification upgrades in existing building stock - differentiating between the residential and commercial sectors. This occupational analysis is intended to be used by communities to inform workforce initiatives (e.g., local training and education programs) as they implement building energy efficiency and electrification projects at scale.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗
NASA NTRS2000

NASA Space Technology Can Improve Soldier Health, Performance and Safety

One of the primary goals of NASA Life Sciences research is '... to enable a permanent human presence in space.' To meet this goal, NASA is creating alternative protocols designed to evaluate and test countermeasures that will account for and correct the environmental effects of space flight on crewmembers health, safety, and operational performance. NASA investigators have previously evaluated the effects of long-duration space flight on physiology and performance of cosmonauts aboard the MIR space station. They also initiated tests of a countermeasure, Autogenic-Feedback Training Exercise (AFTE) designed to prevent and/or correct adverse effects, i.e., facilitate adaptation to space and re-adaptation to Earth. AFTE is a six-hour physiological training program that has proven to be a highly efficient and effective method for enabling people to monitor and voluntarily control a range of their own physiological responses, thereby minimizing adverse reactions to environmental stress. However, because of limited opportunities to test this technology with space flight crews, it is essential to find operational or 'real world' environments in which to validate the efficacy of this approach.

Cowings, Patricia S.↗
NASA NTRS2024

A Koopman-based residual modeling approach for the control of a soft robot arm

Soft robots are challenging to model and control due to their poorly defined kinematics and nonlinear dynamics. Recently, Koopman operator theory has been shown capable of constructing control-oriented soft robot models from data. However, building these models requires extensive data collection and they do not necessarily generalize well outside of the training observations. This paper presents a more data-efficient and generalizable approach to soft robot modeling that first identifies a physics-based Koopman model then supplements it with a data-driven residual Koopman model. The resulting combined model is linear and thus compatible with real-time model-based control techniques such as Model Predictive Control (MPC). The efficacy of the approach is demonstrated on several simulated systems and on a real soft robot arm, where it is shown to generate models that are more accurate than purely physics-based models and require less data to construct than purely data-driven models. Using a model-based controller, the soft arm is able to successfully track end effector trajectories, perform a pick-and-place task, and write on a dry-erase board, showcasing the applicability of this framework to increase the capabilities of soft robotic systems.

Daniel Bruder↗
DOE OSTI2020

Leveraging 3D CAD and Virtual Reality in Design of the Calcine Disposition Project - 20512

Fluor Idaho's Calcine Disposition Project (CDP) is leveraging the power of three-dimensional (3D) visualization, light detection and ranging (Lidar), and virtual reality to improve designs and worker safety. With the use of computer-aided design (CAD) modeling software, the CDP has created 3D models of the facility and systems to aid in the design, development, and operation of its retrieval and transfer system. This by itself is nothing new and has been done in industry for years. The CDP project however is taking this a few steps farther with the use of Lidar and virtual reality software. The CRP is tasked with removing a radioactive granular material from stainless steel bins located in concrete vaults that were constructed in the late 1950's and early 1960's. While construction drawings are available, it is not certain that the drawings are as-built or how accurate they truly are. Part of the project requires precise placement of equipment on to bins around multiple pipe, electrical lines, and bin stiffeners. Relying solely on the existing drawings and models created from these drawings is highly risky and prone to failure. The first step to determine the precise location and orientation of all obstructions in the vault the project will use a Lidar system to scan the vault structure and precisely locate all vault and bin components. Once this information is obtained, the data will be integrated into the CAD model. The model will then be verified, and a true as-built model developed. Precise component location, size, and orientation will be used for final design and placement of retrieval components. The second step will be to take the verified 3D model and scan data and import it into virtual reality software creating a virtual world. While some work on the project requires all work to be completed remotely due to high radiation fields, some areas have low enough radiation levels that personnel can enter and perform work. With the use of a virtual reality headset, operators and technicians will be able to enter our virtual world and become familiar with the surroundings and perform work prior to ever entering the radiation environment. This training is expected to pay dividends by improving worker efficiency, reducing errors, and improving confidence that the work can be performed as expected. It will play a key aspect in as low as reasonably achievable (ALARA) principles. An additional aspect in using the Lidar scans and virtual environment, is that it will give engineers and managers an opportunity to revisit the facility, especially radiation areas, at any time to obtain design information and measurements that may be costly or impossible to obtain once the facility transitions into radiation operations. (authors)

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗
DOE OSTI2024

Reduced-Order Modeling of Multigroup Neutron Cross Sections for High-Temperature Gas-cooled Reactors

Abstract – Deterministic neutronics calculations rely on multigroup neutron cross section libraries, which consist of databases of tabulated values, used to calculate the neutron cross sections through multivariate linear interpolation. However, interpolation of the multidimensional cross section data becomes memory inefficient and time consuming as the number of tabulations increases, significantly slowing down the neutronics calculation, especially in the case of microscopic cross section libraries where every isotope (on the order of hundreds) has its own set of specific reactions and cross sections. In order to address this challenge, this work constructs efficient and robust reduced-order models (ROMs) of the multi-group cross sections to support the Griffin simulation of high-temperature gas-cooled reactors (HTGRs). The first part of the study investigates the linearity of the multigroup cross section data across isotopes, reaction types, and energy groups on pre-generated datasets for the purpose of dimensionality reduction. Secondly, a down-selection of ROM techniques is presented on representative classical machine learning (ML) techniques, including variants of linear regression, kernel-based methods, tree-based algorithms, and artificial neural networks. The selection criteria jointly consider the memory efficiency, predictive accuracy, prediction speed, and scalability in comparison to the multidimensional interpolation. Among all the ML techniques, deep neural networks (DNNs) have proven to be the best selection with sufficient accuracy, high robustness, good memory efficiency, great scalability, and superior flexibility. DNNs have been trained for all isotopes in this work and systematic Griffin testing is ongoing to ensure the feasibility of this ROM technique for predicting cross section and reducing memory requirements without a significant sacrifice in computational performance.

42 - ENGINEERING↗
DOE OSTI2024

Advanced Cross Section Library Generation using Reduced Order Models

Deterministic neutronics calculations rely on multigroup neutron cross section libraries, which consist of databases of tabulated values, used to calculate the neutron cross sections through multivariate linear interpolation. However, interpolation of the multidimensional cross section data becomes memory inefficient and time consuming as the number of tabulations increases, significantly slowing down the neutronics calculation, especially in the case of microscopic cross section libraries where every isotope (on the order of hundreds) has its own set of specific reactions and cross sections. In order to address this challenge, this work constructs efficient and robust reduced-order models (ROMs) of the multi-group cross sections to support the Griffin simulation of high-temperature gas-cooled reactors (HTGRs). The first part of the study investigates the linearity of the multigroup cross section data across isotopes, reaction types, and energy groups on pre-generated datasets for the purpose of dimensionality reduction. Secondly, a down-selection of ROM techniques is presented on representative classical machine learning (ML) techniques, including variants of linear regression, kernel-based methods, tree-based algorithms, and artificial neural networks. The selection criteria jointly consider the memory efficiency, predictive accuracy, prediction speed, and scalability in comparison to the multidimensional interpolation. Among all the ML techniques, deep neural networks (DNNs) have proven to be the best selection with sufficient accuracy, high robustness, good memory efficiency, great scalability, and superior flexibility. DNNs have been trained for all isotopes in this work and systematic Griffin testing is ongoing to ensure the feasibility of this ROM technique for predicting cross section and reducing memory requirements without a significant sacrifice in computational performance.

42 - ENGINEERING↗
NASA NTRS1993

Influence of simulated microgravity on the VO2 max of nontrained and trained rats

Head-down suspension (HDS) of rats has evolved as a useful model for the simulation of a microgravity environment. Previous HDS experiments with rats have shown an impaired capacity to perform aerobic exercise as demonstrated by reductions in maximum oxygen consumption (VO2 max), treadmill run time (RT), and mechanical efficiency (ME) of treadmill running at submaximal conditions. To determine whether endurance training (TR) before HDS would modify exercise performance, male Sprague-Dawley rats were assigned to nontrained (NT) or TR groups for 6 wk and exposed to HDS or cage control (CC) conditions for 29 days. The rats were tested for VO2 max, RT, and ME before treatment and on days 7, 14, 21, and 28. In addition, water and electrolyte excretion was measured on days 1 and 21 of the experimental period. Before HDS, the TR rats had significantly higher measures of VO2 max (15%) and RT (22%) than the NT rats. On day 28, HDS was associated with significant reductions in absolute VO2 max (ml/min) in TR (-30%) and NT (-14%) rats. Relative VO2 max (ml.min-1.kg-1) was significantly reduced in TR (-15%) but not NT rats. Similar reductions in RT occurred in TR (-37%) and NT (-35%) rats by day 28. ME was reduced 22% in both TR and NT rats after 28 days of suspension. HDS elicited diuresis, natriuresis, and kaliuresis in TR rats after 21 days but not after 24 h. In contrast, HDS-NT rats exhibited no diuretic, natriuretic, or kaliuretic responses.(ABSTRACT TRUNCATED AT 250 WORDS).

Non-NASA Center↗
DOE OSTI2025

Machine learning-accelerated path integral molecular dynamics simulations of reactive organic electrolytes

Hydrogen bonded electrolytes that exhibit accelerated proton transport via sequential reactive hops have drawn interest for their promise in clean energy applications. Molecular dynamics simulations of these electrolytes offer the opportunity to uncover microscopic mechanistic details that could be used to design and tune the properties of candidate electrolyte technologies. However, accurately modeling the proton transfer reactions and transport properties that give rise to high charge conductivites in these electrolytes proves computationally challenging because of the need to perform lengthy condensed phase simulations, treating both the electronic and nuclear degrees of freedom quantum mechanically. In this paper, we demonstrate that such a modeling task can be efficiently achieved with the use of density functional theory (DFT)-trained machine learning potentials (MLP) to accelerate path integral molecular dynamics (PIMD) simulations. We highlight the practical utility of this approach by using it to benchmark how closely PIMD simulations employing different DFT exchange–correlation functionals reproduce the composition-dependent densities, diffusion coefficients, and electrical conductivities of mixtures consisting of imidazole and levulinic acid. Even with the speedup afforded by our MLPs, PIMD simulations remain quite expensive. Furthermore, in order to render PIMD more computationally tractable, we introduce and benchmark the accuracy of a ring polymer contraction approach that leverages a computationally efficient short-range MLP to accelerate our PIMD simulations by an additional factor of four.

Chemical bonding↗
DOE OSTI2024

Towards provably efficient quantum algorithms for large-scale machine-learning models

Large machine learning models are revolutionary technologies of artificial intelligence whose bottlenecks include huge computational expenses, power, and time used both in the pre-training and fine-tuning process. In this work, we show that fault-tolerant quantum computing could possibly provide provably efficient resolutions for generic (stochastic) gradient descent algorithms, scaling as $\mathcal{O}$(T 2 x polylog($n$)), where n is the size of the models and T is the number of iterations in the training, as long as the models are both sufficiently dissipative and sparse, with small learning rates. Based on earlier efficient quantum algorithms for dissipative differential equations, we find and prove that similar algorithms work for (stochastic) gradient descent, the primary algorithm for machine learning. In practice, we benchmark instances of large machine learning models from 7 million to 103 million parameters. We find that, in the context of sparse training, a quantum enhancement is possible at the early stage of learning after model pruning, motivating a sparse parameter download and re-upload scheme. Our work shows solidly that fault-tolerant quantum algorithms could potentially contribute to most state-of-the-art, large-scale machine-learning problems.

97 MATHEMATICS AND COMPUTING↗