Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “learning rate”

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 775 records · Page 43
All sourcesNASA NTRSDOE OSTI
DOE OSTI2024

Real-time charged track reconstruction for CLAS12

Abstract This paper presents the results of charged particle track reconstruction in CLAS12 using artificial intelligence. In our approach, we use machine learning algorithms to reconstruct tracks, including their momentum and direction, with high accuracy from raw hits of the CLAS12 drift chambers. The reconstruction is performed in real-time, with the rate of data acquisition, and allows for the identification of event topologies in real-time. This approach revolutionizes the Nuclear Physics experiments' data processing, allowing us to identify and categorize the experimental data on the fly, and will lead to a significant reduction in experiment data processing. It can also be used in streaming readout applications leading to more efficient data acquisition and post-processing.

Instruments & Instrumentation↗
DOE OSTI2024

End-to-end codesign of Hessian-aware quantized neural networks for FPGAs

Here, we develop an end-to-end workflow for the training and implementation of co-designed neural networks (NNs) for efficient field-programmable gate array (FPGA) hardware. Our approach leverages Hessian-aware quantization of NNs, the Quantized Open Neural Network Exchange intermediate representation, and the hls4ml tool flow for transpiling NNs into FPGA firmware. This makes efficient NN implementations in hardware accessible to nonexperts in a single open sourced workflow that can be deployed for real-time machine-learning applications in a wide range of scientific and industrial settings. We demonstrate the workflow in a particle physics application involving trigger decisions that must operate at the 40-MHz collision rate of the CERN Large Hadron Collider (LHC). Given the high collision rate, all data processing must be implemented on FPGA hardware within the strict area and latency requirements. Based on these constraints, we implement an optimized mixed-precision NN classifier for high-momentum particle jets in simulated LHC proton-proton collisions.

47 OTHER INSTRUMENTATION↗
DOE OSTI2020

Advanced Electrochemical Separations of Actinide/Fission Products via the Control of Nucleation and Growth of Electrodeposits. Final Report

This is the final report to the NEUP regarding the “Advanced Electrochemical Separations of Actinide/Fission products via the Control of Nucleation and Growth of Electrodeposits” project. This report describes the project activities conducted during the period October 01, 2016 through September 30, 2019. The report includes a brief statement of the project objectives and approach, followed by the details of the performed research activities and suggestion for the future work. It is well known that according to current knowledge the pyroprocessing of spent nuclear fuels invariably leads to the dendritic form of electrodeposits, which from technological point of view is undesirable for many reasons. The aim of this research was to study and preclude the formation of electrodeposits with dendritic morphology (of essential importance to FC-1 Program). This project is characterized with four novel study categories: (1) exploring electrodeposit morphology control by controlling the electrochemical deposition reaction rate, (2) the use of rotating disk electrode technique to distinguish the electrodeposition control modes (reaction rate vs. diffusion), (3) morphology studies of lanthanides (La) and actinides (U) in characteristic stages of nucleation and growth, (4) exploring the need for separation of electrochemical cell in compartments for cathodic and anodic reactions. All these study categories were applied on two distinct electrolyte systems (a) high temperature ionic liquids, i.e. molten LiCl-KCl salts, and (b) room temperature ionic liquids (compounds exclusively made of organic cations, and organic/inorganic anions). Because the quality (dense and smooth) of electrodeposits is a direct function of the electrochemical reduction rate (slower the better), the major expectation was that addition of organic molecules would provide the desired slowing down of electroreduction effect, avoiding the undesired dependence on mass transfer. What was learned is summarized in seven key areas.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗
DOE OSTI2022

LLNL 5-2428: Fracture Permeability and Seismic Slip Behavior

Our goal is to develop, apply and validate a holistic thermal, hydrologic, mechanical, and chemical (THMC) workflow that also includes evaluation of induced seismic slip in EGS reservoirs. We will integrate experimental and modelling approaches to reduce parameteruncertainty and better predict and mitigate seismic hazard at Utah FORGE and future EGS sites.We propose a novel approach that incorporates 3D physics-based Earthquake simulations in THMC models, herein referred to as “THMC+E” models. This capability will enable improvedengineering decisions at Utah FORGE and move EGS operations toward repeatable, robust, economically viable, and socially accepted development. Utah FORGE management and future EGS operators may employ results of THMC+E models for decision making purposes throughout the lifetime of the field operation. For example, before production well installation, our THMC+E models will predict circulation scenarios and related seismic hazard for a suite ofpossible well locations and flow rates, thus enabling evaluation of optimal production well placement. Such efforts will be conducted throughout the lifetime of the project, whereby additional laboratory experiments will constrain key model parameters and machine learning (ML) will reduce the size of the parameter space and the associated uncertainty. THMC+E simulations will enable exploration various circumstances that may hinder EGS success and develop mitigation strategies.

58 GEOSCIENCES↗
DOE OSTI2026

Automating Detection and Diagnosis of Faults, Failures, and Underperformance in PV Plants

The project developed hybrid physics-based and machine-learning methods for near-real-time detection of balance-of-system faults (e.g., string, combiner, and tracker outages) in utility-scale Photovoltaic plants, achieving over 50% true positive rates with under 10% false positives and significantly reducing engineering setup time. In the extended phase, the scope expanded to plant-level underperformance analysis and industry benchmarking through the SUPER.epri.com platform. SUPER standardizes data processing and performance metrics across more than 9 GWac and 120+ plants, enabling robust comparisons and insights into loss rates, inverter downtime, and capacity degradation.

14 SOLAR ENERGY↗
NASA NTRSDate not supplied

Expanding NeMO-Net Machine Learning Capabilities for Citizen Science

NASA NeMO-Net, the neural multi-modal observation and training network for global coral reef assessment, is an open-source deep convolutional neural network and interactive active learning training software aiming to accurately assess the present and past dynamics of coral reef ecosystems through determination of percent living cover and morphology as well as mapping of spatial distribution. We present an interactive citizen science video game, released this April, for desktop and iOS devices where users interactively label morphology classifications over mm-scale 3D coral reef imagery captured using diver photomosaic imagery, the UAV enabled NASA FluidCam instrument, and satellite datasets. To date, the application has had over 40,000 downloads and over 60,000 unique coral reef classifications, each filtered through a user-based rating and expert evaluation system. We also present results from NeMO-Net’s convolutional neural network (CNN) models used to semantically segment 2D satellite imagery as well as projections of 3D coral reconstructions using user input data as training datasets. Fusing datasets using machine learning from multiple remote sensing platforms presents novel methodologies for assessing the health of coral ecosystems, which are critically endangered by a changing climate. In partnering with Mission Blue, the National Oceanic and Atmospheric Administration (NOAA), and the Living Oceans Foundation (LOF), NeMO-Net leverages an international consortium of subject matter experts to provide both proper training for citizen scientists and the generation of a labeled datasets to ingest into machine learning algorithms for global coral reef identification.

NeMO-Net↗
NASA NTRS2011

Oxygen Uptake Responses to Submaximal Exercise Loads Do Not Change During Long-Duration Space Flight

In previous publications we have reported that the heart rate (HR) responses to graded submaximal exercise tests are elevated during long-duration International Space Station (ISS) flights. Furthermore, the elevation in HR appears greater earlier, rather than later, during the missions. A potential confounder in the interpretation of HR results from graded exercise tests on ISS is that the cycle ergometer used (CEVIS) is vibration-isolated from the station structure. This feature causes the CEVIS assembly to sway slightly during its use and debriefing comments by some crewmembers indicate that there is a "learning curve" associated with CEVIS use. Therefore, one could not exclude the possibility that the elevated HRs experienced in the early stages of ISS missions were related to a lowered metabolic efficiency of CEVIS exercise that would raise the submaximal oxygen uptake (VO2) associated with graded exercise testing work rates.

Moore, Alan D., Jr.↗
DOE OSTI2020

Acoustic Energy Release During the Laboratory Seismic Cycle: Insights on Laboratory Earthquake Precursors and Prediction

Abstract Machine learning can predict the timing and magnitude of laboratory earthquakes using statistics of acoustic emissions. The evolution of acoustic energy is critical for lab earthquake prediction; however, the connections between acoustic energy and fault zone processes leading to failure are poorly understood. Here, we document in detail the temporal evolution of acoustic energy during the laboratory seismic cycle. We report on friction experiments for a range of shearing velocities, normal stresses, and granular particle sizes. Acoustic emission data are recorded continuously throughout shear using broadband piezo‐ceramic sensors. The coseismic acoustic energy release scales directly with stress drop and is consistent with concepts of frictional contact mechanics and time‐dependent fault healing. Experiments conducted with larger grains (10.5 μm) show that the temporal evolution of acoustic energy scales directly with fault slip rate. In particular, the acoustic energy is low when the fault is locked and increases to a maximum during coseismic failure. Data from traditional slide‐hold‐slide friction tests confirm that acoustic energy release is closely linked to fault slip rate. Furthermore, variations in the true contact area of fault zone particles play a key role in the generation of acoustic energy. Our data show that acoustic radiation is related primarily to breaking/sliding of frictional contact junctions, which suggests that machine learning‐based laboratory earthquake prediction derives from frictional weakening processes that begin very early in the seismic cycle and well before macroscopic failure.

58 GEOSCIENCES↗
DOE OSTI2025

Universal energy-speed-accuracy trade-offs in driven nonequilibrium systems

The connection between measure theoretic optimal transport and dissipative nonequilibrium dynamics provides a language for quantifying nonequilibrium control costs, leading to a collection of thermodynamic speed limits, which rely on the assumption that the target probability distribution is perfectly realized. This is almost never the case in experiments or numerical simulations, so here we address the situation in which the external controller is imperfect. We obtain a lower bound for the dissipated work in generic nonequilibrium control problems that (1) is asymptotically tight and (2) matches the thermodynamic speed limit in the case of optimal driving. Along with analytically solvable examples, we refine this imperfect driving notion to systems in which the controlled degrees of freedom are slow relative to the nonequilibrium relaxation rate, and identify independent energy contributions from fast and slow degrees of freedom. Furthermore, we develop a strategy for optimizing minimally dissipative protocols based on optimal transport flow matching, a generative machine learning technique. Furthermore, this latter approach ensures the scalability of both the theoretical and computational framework we put forth. Crucially, we demonstrate that we can compute the terms in our bound numerically using efficient algorithms from the computational optimal transport literature and that the protocols we learn saturate the bound.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2021

Improving Detection of Disease Re-emergence Using a Web-Based Tool (RED Alert): Design and Case Analysis Study

Background: Currently, the identification of infectious disease re-emergence is performed without describing specific quantitative criteria that can be used to identify re-emergence events consistently. This practice may lead to ineffective mitigation. In addition, identification of factors contributing to local disease re-emergence and assessment of global disease re-emergence require access to data about disease incidence and a large number of factors at the local level for the entire world. This paper presents Re-emerging Disease Alert (RED Alert), a web-based tool designed to help public health officials detect and understand infectious disease re-emergence. Objective: Our objective is to bring together a variety of disease-related data and analytics needed to help public health analysts answer the following 3 primary questions for detecting and understanding disease re-emergence: Is there a potential disease re-emergence at the local (country) level? What are the potential contributing factors for this re-emergence? Is there a potential for global re-emergence? Methods: We collected and cleaned disease-related data (eg, case counts, vaccination rates, and indicators related to disease transmission) from several data sources including the World Health Organization (WHO), Pan American Health Organization (PAHO), World Bank, and Gideon. We combined these data with machine learning and visual analytics into a tool called RED Alert to detect re-emergence for the following 4 diseases: measles, cholera, dengue, and yellow fever. We evaluated the performance of the machine learning models for re-emergence detection and reviewed the output of the tool through a number of case studies. Results: Our supervised learning models were able to identify 82%-90% of the local re-emergence events, although with 18%-31% (except 46% for dengue) false positives. This is consistent with our goal of identifying all possible re-emergences while allowing some false positives. The review of the web-based tool through case studies showed that local re-emergence detection was possible and that the tool provided actionable information about potential factors contributing to the local disease re-emergence and trends in global disease re-emergence. Conclusions: To the best of our knowledge, this is the first tool that focuses specifically on disease re-emergence and addresses the important challenges mentioned above.

60 APPLIED LIFE SCIENCES↗
NASA NTRS1999

Pharmacological Inhibitors of the Proteosome in Atrophying Muscles

It is now clear that the marked loss of muscle mass that occurs with disuse, denervation or in many systemic diseases (cancer cachexia, sepsis, acidosis, various endocrine disorders) is due primarily to accelerated degradation of muscle proteins, especially myofibrillar components. Recent work primarily in Dr. Goldberg's laboratory had suggested that in these diverse conditions, the enhancement of muscle proteolysis results mainly from activation of the Ub-proteasome degradative pathway. In various experimental models of atrophy, rat muscles show a common series of changes indicative of activation of this pathway, including increases in MRNA for Ub and proteasome subunits, content of ubiquitinated proteins, and sensitivity to inhibitors of the proteasome. In order to understand the muscle atrophy seen in weightlessness, Dr. Goldberg's laboratory is collaborating with Dr. Baldwin in studies to define the changes in these parameters upon hind-limb suspension. Related experiments will explore the effects on this degradative system of exercise regimens and also of glucocorticoids, which are known to rise in space personnel and to promote muscle, especially in inactive muscles. The main goals will be: (A) to define the enzymatic changes leading to enhanced activity of the Ub-proteasome pathway in inactive muscles upon hind-limb suspension, and the effects on this system of exposure to glucocorticoids or exercise; and (B) to learn whether inhibitors of the Ub-proteasome pathway may be useful in retarding the excessive proteolysis in atrophying muscles. Using muscle extracts, Dr. Goldberg's group hopes to define the rate-limiting, enzymatic changes that lead to the accelerated Ub-conjugation and protein degradation. They have recently developed cell-free preparations from atrophying rat muscles, in which Ub-conjugation to muscle proteins is increased above control levels. Because these new preparations seem to reproduce the changes occurring in vivo, they will analyze in depth extracts from normal and atrophying muscles to compare the activities of the Ub-activating enzyme (El), the various LTh-carrier proteins (E2s), and Ub-protein ligases (E3s). Recent studies of other types of muscle wasting -suggest a very important role in muscle proteolysis of certain ubiquitination enzymes, E214k and E3-alpha(i.e. components of the "N-end pathway"). Future studies will focus in understanding their role and test whether they are in fact critical for muscle atrophy in vivo. Since weightlessness leads to a specific loss of contractile proteins and to a switching of myosin isotypes, Dr. Goldberg's group will attempt to identify the ubiquitination enzymes specifically involved in myosin degradation both in normal muscle and after hind-limb suspension.

Goldberg, Alfred↗
NASA NTRS1980

Safe Handling Practices

In 1977 Compugraphic Corporation was experiencing an unacceptable failure rate on microelectronic chips. Company engineers suspected that static electricity was causing the trouble because some electronic components are highly susceptible to damage by electrostatic charge. From a NASA Tech Brief, they learned that Rockwell International had prepared a report on safe handling practices for electronic components. NASA provided a Technical Support Package detailing 50 safe handling procedures affecting workers, work areas, equipment and packaging materials. Where poor practices were discovered, re-education of employees and other corrective measures were undertaken.

Source record↗
NASA NTRS2011

Ground Processing Affordability for Space Vehicles

Launch vehicles and most of their payloads spend the majority of their time on the ground. The cost of ground operations is very high. So, why so often is so little attention given to ground processing during development? The current global space industry and economic environment are driving more need for efficiencies to save time and money. Affordability and sustainability are more important now than ever. We can not continue to treat space vehicles as mere science projects. More RLV's (Reusable Launch Vehicles) are being developed for the gains of reusability which are not available for ELV's (Expendable Launch Vehicles). More human-rated vehicles are being developed, with the retirement of the Space Shuttles, and for a new global space race, yet these cost more than the many unmanned vehicles of today. We can learn many lessons on affordability from RLV's. DFO (Design for Operations) considers ground operations during design, development, and manufacturing-before the first flight. This is often minimized for space vehicles, but is very important. Vehicles are designed for launch and mission operations. You will not be able to do it again if it is too slow or costly to get there. Many times, technology changes faster than space products such that what is launched includes outdated features, thus reducing competitiveness. Ground operations must be considered for the full product Lifecycle, from concept to retirement. Once manufactured, launch vehicles along with their payloads and launch systems require a long path of processing before launch. Initial assembly and testing always discover problems to address. A solid integration program is essential to minimize these impacts, as was seen in the Constellation Ares I-X test rocket. For RLV's, landing/recovery and post-flight turnaround activities are performed. Multi-use vehicles require reconfiguration. MRO (Maintenance, Repair, and Overhaul) must be well-planned--- even for the unplanned problems. Defect limits and standard repairs need to be in-place as well as easily added. Many routine inspections and maintenance can be like an aircraft overhaul. Modifications and technology upgrades should be expected. Another factor affecting ground operations efficiency is trending. It is essential for RLV's, and also useful for ELV's which fly the same or similar models again. Good data analysis of technical and processing performance will determine fixes and improvements needed for safety, design, and future processing. Collecting such data on new or low-frequency vehicles is a challenge. Lessons can be learned from the Space Shuttle, or even the Concorde aircraft. For all of the above topics, efficient business systems must be established for comprehensive program management and good throughput. Drawings, specifications, and manuals for an entire launch vehicle are often in different formats from multiple vendors, plus they have proprietary constraints. Nonetheless, the integration team must ensure that all data needed is compatible and visible to each appropriate team member. Ground processing systems for scheduling, tracking, problem resolution, etc. must be well laid-out. The balance between COTS (commercial off the shelf) and custom software is difficult. Multiple customers, vendors, launch sites, and landing sites add to the complexity of efficient IT (Information Technology) tools.

Ingalls, John↗
NASA NTRS2014

Application of Fault Management Theory to the Quantitative Selection of a Launch Vehicle Abort Trigger Suite

The theory of System Health Management (SHM) and of its operational subset Fault Management (FM) states that FM is implemented as a "meta" control loop, known as an FM Control Loop (FMCL). The FMCL detects that all or part of a system is now failed, or in the future will fail (that is, cannot be controlled within acceptable limits to achieve its objectives), and takes a control action (a response) to return the system to a controllable state. In terms of control theory, the effectiveness of each FMCL is estimated based on its ability to correctly estimate the system state, and on the speed of its response to the current or impending failure effects. This paper describes how this theory has been successfully applied on the National Aeronautics and Space Administration's (NASA) Space Launch System (SLS) Program to quantitatively estimate the effectiveness of proposed abort triggers so as to select the most effective suite to protect the astronauts from catastrophic failure of the SLS. The premise behind this process is to be able to quantitatively provide the value versus risk trade‐off for any given abort trigger, allowing decision makers to make more informed decisions. All current and planned crewed launch vehicles have some form of vehicle health management system integrated with an emergency launch abort system to ensure crew safety. While the design can vary, the underlying principle is the same: detect imminent catastrophic vehicle failure, initiate launch abort, and extract the crew to safety. Abort triggers are the detection mechanisms that identify that a catastrophic launch vehicle failure is occurring or is imminent and cause the initiation of a notification to the crew vehicle that the escape system must be activated. While ensuring that the abort triggers provide this function, designers must also ensure that the abort triggers do not signal that a catastrophic failure is imminent when in fact the launch vehicle can successfully achieve orbit. That is, the abort triggers must have low false negative rates to be sure that real crew‐threatening failures are detected, and also low false positive rates to ensure that the crew does not abort from non‐crew‐threatening launch vehicle behaviors. The analysis process described in this paper is a compilation of over six years of lessons learned and refinements from experiences developing abort triggers for NASA's Constellation Program (Ares I Project) and the SLS Program, as well as the simultaneous development of SHM/FM theory. The paper will describe the abort analysis concepts and process, developed in conjunction with SLS Safety and Mission Assurance (S&MA) to define a common set of mission phase, failure scenario, and Loss of Mission Environment (LOME) combinations upon which the SLS Loss of Mission (LOM) Probabilistic Risk Assessment (PRA) models are built. This abort analysis also requires strong coordination with the Multi‐Purpose Crew Vehicle (MPCV) and SLS Structures and Environments (STE) to formulate a series of abortability tables that encapsulate explosion dynamics over the ascent mission phase. The design and assessment of abort conditions and triggers to estimate their Loss of Crew (LOC) Benefits also requires in‐depth integration with other groups, including Avionics, Guidance, Navigation and Control(GN&C), the Crew Office, Mission Operations, and Ground Systems. The outputs of this analysis are a critical input to SLS S&MA's LOC PRA models. The process described here may well be the first full quantitative application of SHM/FM theory to the selection of a sensor suite for any aerospace system.

Lo, Yunnhon↗
NASA NTRS2023

Electronic Nose Development and Preliminary Human Breath Testing for Rapid, Non-Invasive COVID-19 Detection

We adapted an existing, spaceflight-proven, robust “electronic nose” (E-Nose) that uses an array of electrical resistivity-based nanosensors mimicking aspects of mammalian olfaction to conduct on-site, rapid screening for COVID-19 infection by measuring the pattern of sensor responses to volatile organic compounds (VOCs) in exhaled human breath. We built and tested multiple copies of a hand-held prototype E-Nose sensor system, composed of 64 chemically sensitive nanomaterial sensing elements tailored to COVID-19 VOC detection; data acquisition electronics; a smart tablet with software (App) for sensor control, data acquisition and display; and a sampling fixture to capture exhaled breath samples and deliver them to the sensor array inside the E-Nose. The sensing elements detect the combination of VOCs typical in breath at parts-per-billion (ppb) levels, with repeatability of 0.02% and reproducibility of 1.2%; the measurement electronics in the E-Nose provide measurement accuracy and signal-to-noise ratios comparable to benchtop instrumentation. Preliminary clinical testing at Stanford Medicine with 63 participants, their COVID-19-positive or COVID-19-negative status determined by concomitant RT-PCR, discriminated between these two categories of human breath with a 79% correct identification rate using “leave-one-out” training-and-analysis methods. Analyzing the E-Nose response in conjunction with body temperature and other non-invasive symptom screening using advanced machine learning methods, with a much larger database of responses from a wider swath of the population, is expected to provide more accurate on-the-spot answers. Additional clinical testing, design refinement, and a mass manufacturing approach are the main steps toward deploying this technology to rapidly screen for active infection in clinics and hospitals, public and commercial venues, or at home.

COVID-19↗
DOE OSTI2024

Improved deep learning prediction of antigen–antibody interactions

Identifying antibodies that neutralize specific antigens is crucial for developing effective immunotherapies, but this task remains challenging for many target antigens. The rise of deep learning–based computational approaches presents a promising avenue to address this challenge. Here, we assess the performance of a deep learning approach through two benchmark tests aimed at predicting antibodies for the receptor-binding domain of the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) spike protein. Three different strategies for constructing input sequence alignments are employed for predicting structural models of antigen–antibody complexes. In our initial testing set, which comprises known experimental structures, these strategies collectively yield a significant top-ranked prediction for 61% of cases and a success rate of 47%. Notably, one strategy that utilizes the sequences of known antigen binders outperforms the other two, achieving a precision of 90% in a subsequent test set of ~1,000 antibodies, balanced between true and control antibodies for the antigen, albeit with a lower recall of 25%. Our results underscore the potential of integrating deep learning methods with single B cell sequencing techniques to enhance the prediction accuracy of antigen–antibody interactions.

Science & Technology - Other Topics↗
NASA NTRS2023

Machine Learning Framework for Hazard Extraction and Analysis of Trends (HEAT) in Wildfire Response

This research proposes a natural language processing enabled risk analysis framework, named Hazard Extraction andAnalysis of Trends (HEAT), and applies the framework to the ICS-209-PLUS data set of wildfire incident responseforms. The HEAT framework produces safety- and risk- relevant analyses, consisting of: (1) a set of hazards extractedfrom text data, (2) a primary analysis using hazard-relevant metrics, such as rate and severity, to form an FMEA-styletable and risk matrix, (3) a time series analysis of metric trends, and (4) a secondary analysis examining potentialpredictors for hazards. Results from HEAT provide quantitative risk-relevant information for high-level hazards doc-umented in existing-state operations. Because of the generalizability of the steps and limited data requirements, HEATcan be applied to any dataset containing narrative text, thus providing a framework for data-driven machine learning-enabled quantitative risk analysis across a variety of domains. To demonstrate HEAT in a case study, we apply theframework to the ICS-209-PLUS dataset of wildland fire incident response forms. Hazards identified in wildfire re-sponse arise from environmental conditions, the mission, and the wildland urban interface. The resulting risk matrixidentifies evacuations as high-risk hazards, while all other identified hazards are medium or serious risk.

natural language processing↗
DOE OSTI2025

Fair Concurrent Training of Multiple Models in Federated Learning

Federated learning (FL) enables collaborative learning across multiple clients. In most FL work, all clients train a single learning task. However, the recent proliferation of FL applications may increasingly require multiple FL tasks to be trained simultaneously, sharing clients’ computing resources, which we call Multiple-Model Federated Learning (MMFL). Current MMFL algorithms use naïve average-based client-task allocation schemes that often lead to unfair performance when FL tasks have heterogeneous difficulty levels, as the more difficult tasks may need more client participation to train effectively. Furthermore, in the MMFL setting, we face a further challenge that some clients may prefer training specific tasks to others, and may not even be willing to train other tasks, e.g., due to high computational costs, which may exacerbate unfairness in training outcomes across tasks. We address both challenges by firstly designing FedFairMMFL, a difficulty-aware algorithm that dynamically allocates clients to tasks in each training round, based on the tasks’ current performance levels. We provide guarantees on the resulting task fairness and FedFairMMFL’s convergence rate. We then propose novel auction designs that incentivizes clients to train multiple tasks, so as to fairly distribute clients’ training efforts across the tasks, and extend our convergence guarantees to this setting. Here, we finally evaluate our algorithm with multiple sets of learning tasks on real world datasets, showing that our algorithm improves fairness by improving the final model accuracy and convergence speed of the worst performing tasks, while maintaining the average accuracy across tasks.

Federated learning↗