SEARCH · Search NASA

Results for “AI models”

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 325 records · Page 18

All sources NASA NTRS DOE OSTI

DOE OSTI2025

Agentic AI vs ML-Based Autotuning: A Comparative Study for Loop Reordering Optimization

High Performance Computing (HPC) applications rely heavily on code optimizations to achieve good performance on modern CPU and GPU architectures. Traditional Machine Learning auto-tuning approaches have demonstrated success in exploring high-dimensional spaces, but they often require expensive compile-run evaluations and lack adaptability for large HPC applications. The recent advances in Large Language Models (LLMs) and Agentic AI systems raise intriguing questions about the potential of these approaches to address specific optimization methodologies. This work aims to answer an essential question for the HPC community: “How Agentic AI Systems Compare to Traditional ML Autotuning Techniques?” To address this question, we present a comparative analysis between a traditional ML-based optimization approach and an Agentic AI system, evaluating their respective capabilities and limitations for loop-level optimization. In addition, we introduced a new Agentic AI system named LoopGen-AI using three different Large Language Models: GPT-4.1, Claude 4.0, and Gemini 2.5. A key finding is that LoopGen-AI achieves competitive per-formance with only a few program runs, the reasoning logs from the agents revealed that their decisions rely heavily on the combination of semantic understanding of the target kernel with dynamic feedback from the environment, highlighting a promising new dimension in performance tuning. In contrast, ML-based autotuners focus on statistical exploration, and require orders of magnitude more runs to reach peak performance. Additionally, our analysis shows that prompt engineering, particularly using Persona + Context Manager patterns, significantly impacts the effectiveness of Agentic AI. Our results indicate that while Agentic AI systems are not yet a complete replacement for ML-based autotuners, it can effectively complement traditional methods.

Rosas, Miguel Romero↗

DOE OSTI2024

Artificial Intelligence Transforming Post-Translational Modification Research

Post-Translational Modifications (PTMs) are covalent changes to amino acids that occur after protein synthesis, including covalent modifications on side chains and peptide backbones. Many PTMs profoundly impact cellular and molecular functions and structures, and their significance extends to evolutionary studies as well. In light of these implications, we have explored how artificial intelligence (AI) can be utilized in researching PTMs. Initially, rationales for adopting AI and its advantages in understanding the functions of PTMs are discussed. Then, various deep learning architectures and programs, including recent applications of language models, for predicting PTM sites on proteins and the regulatory functions of these PTMs are compared. Finally, our high-throughput PTM-data-generation pipeline, which formats data suitably for AI training and predictions is described. We hope this review illuminates areas where future AI models on PTMs can be improved, thereby contributing to the field of PTM bioengineering.

59 BASIC BIOLOGICAL SCIENCES↗

DOE OSTI2025

White Paper: Scalable Digital Twin Capabilities for Aging and Surveillance of Engineered Systems

This white paper presents a multi-year initiative to develop practical, secure, and scalable digital twin capabilities for engineered systems in aging and surveillance contexts—an approach pioneered at the National Nuclear Security Administration (NNSA) Lawrence Livermore National Laboratory (LLNL) that maps directly onto the needs and ambitions of the Navy for ship- and fleet-level digital twins. LLNL’s work in building part- and process-level digital twins for advanced manufacturing, with a vision to scale up to entire factory floors and, ultimately, enterprise-wide digital twins, offers an adaptable pathway for the Navy as it seeks to modernize lifecycle management, readiness, and predictive maintenance across ships and fleets. For our application, we integrate physics-based modeling with automated data ingestion, processing, and AI-driven calibration, creating hybrid models that are both interpretable and data responsive. We modernized legacy workflows, established centralized data infrastructure, automated experimental pipelines, and demonstrated end-to-end coupling of accelerated aging data with finite element simulations via optimization and surrogate modeling. The result is a generalizable framework that supports part-level digital twins today and lays the groundwork for future system-level twins suitable for Navy applications.

36 MATERIALS SCIENCE↗

DOE OSTI2025

Advancing AI-Driven Analysis in X-ray Absorption Spectroscopy: Spectral Domain Mapping and Universal Models

In recent years, rapid progress has been made in developing artificial intelligence (AI) and machine learning (ML) methods for X-ray absorption spectroscopy (XAS) analysis. Compared to traditional XAS analysis methods, AI/ML approaches offer dramatic improvements in efficiency and help eliminate human bias. To advance this field, we advocate an AI-driven XAS analysis pipeline that features several interconnected key building blocks: benchmarks, workflows, databases, and AI/ML models. Specifically, we present two case studies for XAS ML. In the first study, we demonstrate the importance of reconciling the discrepancies between simulation and experiment using spectral domain mapping (SDM). Our ML model, which is trained solely on simulated spectra, predicts an incorrect oxidation state trend for Ti atoms in a combinatorial zinc titanate film. After transforming the experimental spectra into a simulation-like representation using SDM, the same model successfully recovers the correct oxidation state trend. In the second study, we explore the development of universal XAS ML models that are trained on the entire periodic table, which enables them to leverage common trends across elements. Looking ahead, we envision that an AI-driven pipeline can unlock the potential of real-time XAS analysis to accelerate scientific discovery.

36 MATERIALS SCIENCE↗

DOE OSTI2021

Artificial Intelligence for Data Center Operations (AI Ops)

HPC data centers such as the one at NREL's ESIF will increasingly need to rely on automation to keep pace with exascale growth in compute capability and to manage and optimize the data center environment and facility resources. Artificial intelligence (AI) and machine learning (ML) approaches provide the means to improve HPC data center efficiency (energy, operational, and managerial efficiency) and resiliency by learning historical trends and training models to operate on real-time data collected from both IT and facilities sources. The goal of coupled improvement of data center resiliency and energy efficiency through automated data collection and AI has led to a multi-year, multi-staged collaboration between NREL and Hewlett-Packard Enterprise's Advanced Technology Group, referred to as Artificial Intelligence for Data Center Operations (AIOps). The extended efforts within the AIOps project include a common goal of building capabilities for an advanced smart facility and demonstration of data collection and AI modeling techniques in the ESIF data center.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2024

Perspectives on AI Architectures and Codesign for Earth System Predictability

Abstract Recently, the U.S. Department of Energy (DOE), Office of Science, Biological and Environmental Research (BER), and Advanced Scientific Computing Research (ASCR) programs organized and held the Artificial Intelligence for Earth System Predictability (AI4ESP) workshop series. From this workshop, a critical conclusion that the DOE BER and ASCR community came to is the requirement to develop a new paradigm for Earth system predictability focused on enabling artificial intelligence (AI) across the field, laboratory, modeling, and analysis activities, called model experimentation (ModEx). BER’s ModEx is an iterative approach that enables process models to generate hypotheses. The developed hypotheses inform field and laboratory efforts to collect measurement and observation data, which are subsequently used to parameterize, drive, and test model (e.g., process based) predictions. A total of 17 technical sessions were held in this AI4ESP workshop series. This paper discusses the topic of the AI Architectures and Codesign session and associated outcomes. The AI Architectures and Codesign session included two invited talks, two plenary discussion panels, and three breakout rooms that covered specific topics, including 1) DOE high-performance computing (HPC) systems, 2) cloud HPC systems, and 3) edge computing and Internet of Things (IoT). We also provide forward-looking ideas and perspectives on potential research in this codesign area that can be achieved by synergies with the other 16 session topics. These ideas include topics such as 1) reimagining codesign, 2) data acquisition to distribution, 3) heterogeneous HPC solutions for integration of AI/ML and other data analytics like uncertainty quantification with Earth system modeling and simulation, and 4) AI-enabled sensor integration into Earth system measurements and observations. Such perspectives are a distinguishing aspect of this paper. Significance Statement This study aims to provide perspectives on AI architectures and codesign approaches for Earth system predictability. Such visionary perspectives are essential because AI-enabled model-data integration has shown promise in improving predictions associated with climate change, perturbations, and extreme events. Our forward-looking ideas guide what is next in codesign to enhance Earth system models, observations, and theory using state-of-the-art and futuristic computational infrastructure.

54 ENVIRONMENTAL SCIENCES↗

DOE OSTI2020

Development of the Intelligent, Preventive Infrared (IR) Inspection System Housed in Hybrid Robotic Platforms

Robots and robotic systems that are designed for inspection, environmental study, and health and safety aid are becoming an increasing necessity. However, there is a number of challenges that accompany robots that are designed for these specific applications. These challenges include: navigating compact, enclosed spaces, travelling over multiple terrains and large obstacles, using the proper sensing and detection methods to assess an environment, and the use of lightweight and durable materials. The robotic platforms currently in development look at all of these challenges and attempt to overcome them. These designs specific use of hybrid robotic platforms, or platform that utilizes soft and rigid materials, allows for a more flexible platform and makes environments more navigable. To further improve the navigation of the platforms and environmental assessment, a novel infrared detection system is housed in the platforms to create a robot that can be used for the applications listed above and more. Objectives: Further develop two types of robotic platforms that utilize additive manufacturing, soft materials, and rigid materials. Continue the development of an intelligent and inhibitory infrared detection system based on an artificial intelligence (AI) algorithm. Continued study and fabrication of active soft materials designed for both sensing and actuation in hybrid robotic systems. Improve additive manufacturing fabrication to design rigid and semi-rigid components for hybrid robotic platforms. Transformable Wheel Robotic Platform: The chassis, wheels, tires and inspection system housing use different additive manufacturing techniques for fabrication. Continued work with additive manufacturing has lead to studies in metal-based printing and modular design and manufacturing. The new platform design with integrated electrical component printed. This will allow integration of the sensor housing onto the platform. Electrical components are being tested for battery life and performance. To improve this performance, such as integration of Lithium Polymer (LiPo) batteries. Snake Robotic Platform: The main focus of the development has centered around liquid-based soft actuators. That act on the principles of electrostatic and hydraulic actuation. A liquid dielectric sits between two compliant electrodes, contained by a flexible polymer shell. The electrodes and film gradually collapse toward each other from one corner of the electrode to the other. When the electrodes and film close together, a majority of the fluid is pushed into the area not covered by an electrode. A thin layer of the liquid dielectric remains between the electrode. The actuators will be stacked to cause large displacement, and move the linkages. The chassis of this platform uses purely additively manufactured linkages. Intelligent, Preventive IR Inspection System: Development of the AI for the system has lead to using a Scikit-Learn which assists in creating predictive models based on Regression, clustering, classification etc. To improve the infrared thermometry for low emissivity sources, work on the fabrication of a tandem photoconductive infrared thermometer was a main focus. Distance-Voltage-Temperature response data has been collected in the range of 7 cm - 100 cm and 200-400 deg. C. Modifications were made to the existing test bench to have a better control over the data. Automated data collection was realized using a Python code and an Arduino controlled stepper motor to increase the sample rate. A protective enclosure has been built around the setup to minimize the effect of the environment on the measurements. To better predict temperature, different AI models are being optimized. The regression model, LARS showed a high accuracy but had convergence issues and only works for the current test set-up. Results: The Transformable Wheel Robot has developed into a more flexible and modular platform. With the improvements to the current work, effort on the tire or soft gripper design been a large focus. The soft grippers will be interchange able to allow for increased performance in identified terrain types. The development of the liquid-based actuators allows the snake robotic platform to achieve the goals of being flexible and able to navigate confined spaces. However, there is room for improvement. Optimization work is currently being done in COMSOL Multiphysics. With the current IR system set-up the LARS model perfectly predicts the data; however, considering the mobility aspect of the project other models will allow for an optimized system. Testing of other model types is currently being done.

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗

DOE OSTI2021

Deep Analysis Net with Causal Embedding for Coal-fired Power Plant Fault Detection and Diagnosis (DANCE4CFDD)

Fault detection and diagnosis is critical to power plant operation to ensure attaining high reliability while reducing operation cost. As more renewable power is introduced to the power grid, traditional fossil power plants take on the extra burden of excessive load cycling to compensate the generation variability from renewable power. Such load cycling will pose more reliability challenges to power plant operation. There are a number of challenges faced by today’s asset health management system in coal- fired (or gas) power plants: 1) high-dimensional nonlinear interaction among multiple time series measurements; 2) high measurement variance induced by operational conditions/modes; 3) variation among asset types and plant configurations; and 4) a small number of faulty events to learn from. To cope with these challenges, today’s fielded asset health management systems rely heavily on manual efforts from domain experts and hand-crafted features or rules based on domain knowledge. Despite its role in plant reliability, such a practice is costly and hinders its scalability and sustainability, particularly when a plant undergoes modifications. The objective of this project is to develop a novel end-to-end AI learning system that is trainable (i.e., the AI representation of a complex system behavior can be directly learned from properly labeled data) for accurate fault detection and root cause analysis. The ability to create a fault detection model directly from time series could alleviate the efforts associated with today’s asset management solution development. In the course of this project, we have achieved the following: Created an AI model development environment incorporating state-of-the-art neural network architectures for rapid model development and evaluation; Developed novel learning strategies for training of fault detection model; Developed special-purpose neural network architecture embedded with variable association graph aiming for better interpretability; Developed a learning strategy to leverage a small number of faulty events for enhanced fault detection capability; Conducted detailed experimental study based on public benchmark datasets and demonstrated the effectiveness of the proposed solution; and Validated the developed system with data from both a coal-fired plant boiler dynamic simulation model and real-world coal-fired power plant covering multiple asset and fault types. Overall, the project attained a technology readiness level of TRL 5 from TRL 2 at the beginning of the project.

20 FOSSIL-FUELED POWER PLANTS↗

DOE OSTI2023

Vehicle trajectory prediction near or at traffic signal

A system and method for determining a predicted trajectory of a human-driven host vehicle as the human-driven host vehicle approaches a traffic signal. The method includes: obtaining a host vehicle-traffic light distance d x and a longitudinal host vehicle speed v x that are each taken when the human-driven host vehicle approaches the traffic signal; obtaining a traffic light signal phase P t and an traffic light signal timing T t ; obtaining a time of day TOD; providing the host vehicle-traffic light distance d x , the longitudinal host vehicle speed v x , the traffic light signal phase P t , the traffic light signal timing T t , and the time of day TOD as input into an artificial intelligence (AI) vehicle trajectory prediction application, wherein the AI vehicle trajectory prediction application implements an AI vehicle trajectory prediction model; and determining the predicted trajectory of the human-driven host vehicle using the AI vehicle trajectory prediction application.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2021

Development of the IES Plug-and-Play Framework

This report discusses the status of the flexible plug-and-play framework development currently ongoing that aims to integrate Modelica/Dymola with the Risk Analysis and Virtual ENvironment (RAVEN) software in terms of both Functional Mock-Up Interface (FMI)/Functional Mock-Up Unit (FMU) construction and repository structures that aim to ease the sharing and simulation of complex dynamic models. This report aims to provide an overview of all the performed activities resolving around the deployment of methods, software infrastructures, guidelines and workflow for the construction and usage of models, encapsulated using the FMI/FMU protocols and standards. In particular, the report is organized in three main macro-subjects, which are connected to each other: - FMI/FMU adaptors for modelica models - HYBRID repository new structure and open-source deployment - RAVEN FMI/FMU exporting capabilities and Artificial Intelligence (AI)-based analysis acceleration. The first part of the report discusses the FMI/FMU adaptors that have been created within the HYBRID repository to allow users to quickly export models, such as FMUs. Several examples are shown that highlight the step-by-step process of converting an existing Modelica model into an FMU for use within the Dymola platform. Simulation results demonstrate that, while minor differences may occur, the overall control, trends, and solution integrity are maintained between standard Modelica simulation and FMU simulation results. However, it is worth noting that, for small systems, the FMU results have a slower simulation time than the Modelica only simulation. Using this process, a company can provide models that contain proprietary information to entities without disclosing any of the information about the model that could be considered business sensitive. Such an ability would allow institutions to bypass the necessity of “whitewashing” data. In the second part of the report, the new structure of the HYBRID repository is discussed with a major focus on the series of updates that has been completed. These updates include the addition of Modelica system-level regression tests and software quality assurance documentation that ensure that modifications to the Modelica models do not alter system-level model results. The third and final part of the report aims to report the work that has been performed for the deployment of methods and workflows for the construction of RAVEN AI-based models compliant with the FMI/FMU standard. Such development represents the key for the deployment of the concept of “Flexible ecosystem” since it allows for the replacement of high-fidelity modelica models (or any other FMI/FMU compliant model) with RAVEN generated AI surrogate models. Overall, extensive work has been completed on developing FMUs and FMIs from existing models, understanding the requirements and limitations of FMUs, and open-sourcing the HYBRID repository with an integrated regression system.

42 ENGINEERING↗

DOE OSTI2021

Development of the IES Plug-and-Play Framework

14 SOLAR ENERGY↗

DOE OSTI2022

Mobility and Energy Improvements Realized through Prediction-based Vehicle Powertrain Control and Traffic Management (Research Performance Final Report)

The goal of this project was to utilize individual vehicle and systems-level transportation big data to develop real-world implementable techniques for energy efficiency. We collected a real world dataset in Fort Collins, CO using technology that is currently available. This dataset was used to (1) create individual vehicle prediction models and emissions models using cutting-edge artificial intelligence (AI) techniques, (2) create traffic prediction models, (3) create boundary condition constraints for optimal trajectory derivation, and (4) develop a novel Mobility Energy Productivity (MEP) model for Fort Collins, CO at a fidelity and flexibility which did not previously exist. Each of these techniques is novel but the most interesting results occurred from the intersection of all techniques. It was found that when optimal vehicle control is combined with optimal traffic light control, that significant energy efficiency improvements that do not compromise travel time are available.

42 ENGINEERING↗

DOE OSTI2020

A non-cooperative meta-modeling game for automated third-party calibrating, validating and falsifying constitutive laws with parallelized adversarial attacks

The evaluation of constitutive models, especially for high-risk and high-regret engineering applications, requires efficient and rigorous third-party calibration, validation and falsification. While there are numerous efforts to develop paradigms and standard procedures to validate models, difficulties may arise due to the sequential, manual, and often biased nature of the commonly adopted calibration and validation processes, thus slowing down data collections, hampering the progress towards discovering new physics, increasing expenses and possibly leading to misinterpretations of the credibility and application ranges of proposed models. This work attempts to introduce concepts from game theory and machine learning techniques to overcome many of these existing difficulties. Here, we introduce an automated meta-modeling game where two competing AI agents systematically generate experimental data to calibrate a given constitutive model and to explore its weakness such that the experiment design and model robustness can be improved through competitions. The two agents automatically search for the Nash equilibrium of the meta-modeling game in an adversarial reinforcement learning framework without human intervention. In particular, a protagonist agent seeks to find the more effective ways to generate data for model calibrations, while an adversary agent tries to find the most devastating test scenarios that expose the weaknesses of the constitutive model calibrated by the protagonist. By capturing all possible design options of the laboratory experiments into a single decision tree, we recast the design of experiments as a game of combinatorial moves that can be resolved through deep reinforcement learning by the two competing players. Our adversarial framework emulates idealized scientific collaborations and competitions among researchers to achieve a better understanding of the application range of the learned material laws and prevent misinterpretations caused by conventional AI-based third-party validation. Numerical examples are given to demonstrate the wide applicability of the proposed meta-modeling game with adversarial attacks on both human-crafted constitutive models and machine learning models.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2022

Hybrid AI-ML and FE-based Digital Twin Predictive Modeling Framework for a PWR Coolant System Components: Updates on Multi-Time-Series-3D-Location Dependent Usages Factor Prediction

The long-term operation (LTO) of nuclear power plants (NPP) beyond their original design life of 40 years can lead to more material damage associated with cyclic fatigue under thermal-mechanical loading cycles and associated long-term exposure of reactor material to the deleterious reactor-coolant environments. However, under this LTO condition, the reactor components can still safely operate but may require more frequent Nondestructive Evaluation (NDE) of reactor components. Requiring frequent NDE inspections may lead to frequent NPP shutdowns which can lead to power outages and additional NDE inspection cost-related economic loss. The economic loss can be minimized by reducing uncertainty in life estimation of safety-critical pressure boundary components and by implementing a more digital approach such as using upcoming digital-twin (DT) technology for predicting the structural states (e.g., time and location dependent inside/outside thickness temperature, stress, strain, plastic deformation, etc.) and associated fatigue life of a component in real time. The DT framework is based on limited experimental data, Artificial-intelligence (AI)-Machine-Learning (ML) and multiphysics-computationalmechanics such as finite element- (FE) based models. Given the real-time thermal-hydraulic process measurements from several existing plant sensors, the overall goal of the DT framework is to predict the cumulative usages factors or equivalent fatigue lives in real time and at any random 3D location of the components. This includes inaccessible locations such as inside the thickness or location of a component. This prediction can be at thousands to millions of 3D point clouds or locations like conventional FE-based models, but without running an FE model in real time.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗

DOE OSTI2024

Integrated edge-to-exascale workflow for real-time steering in neutron scattering experiments

We introduce a computational framework that integrates artificial intelligence (AI), machine learning, and high-performance computing to enable real-time steering of neutron scattering experiments using an edge-to-exascale workflow. Focusing on time-of-flight neutron event data at the Spallation Neutron Source, our approach combines temporal processing of four-dimensional neutron event data with predictive modeling for multidimensional crystallography. At the core of this workflow is the Temporal Fusion Transformer model, which provides voxel-level precision in predicting 3D neutron scattering patterns. The system incorporates edge computing for rapid data preprocessing and exascale computing via the Frontier supercomputer for large-scale AI model training, enabling adaptive, data-driven decisions during experiments. This framework optimizes neutron beam time, improves experimental accuracy, and lays the foundation for automation in neutron scattering. Although real-time experiment steering is still in the proof-of-concept stage, the demonstrated potential of this system offers a substantial reduction in data processing time from hours to minutes via distributed training, and significant improvements in model accuracy, setting the stage for widespread adoption across neutron scattering facilities and more efficient exploration of complex material systems.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2021

Characterization of Extreme Hydroclimate Events in Earth System Models using ML/AI

Focal Area(s): (1) We put forward the concepts of data assimilation enabled by machine learning, AI, and advanced methods including experimental/network design/optimization and unsupervised learning applied to downscale information within Earth System Models (ESMs). (2) We discuss predictive modeling through the use of AI techniques and other tools to design a prediction system comprising of a hierarchy of models (e.g., AI-driven model/component/parameterization selection) to improve the characterization of extreme hydroclimate events in ESMs. The AI-based models will run five-six order of magnitude times faster, yet will provide similar accuracy, allowing us to provide range bounds on uncertainty faster and thus enabling faster extreme event identification. (3) Further, we consider the insight gleaned from complex data (observed/simulated) using AI, big data analytics, and other advanced methods, including explainable AI and physics- or knowledge-guided AI to improve the characterization of extreme hydroclimate events in ESMs.

54 ENVIRONMENTAL SCIENCES↗

DOE OSTI2023

Benchmarking for AI for Science

AI has been instrumental for recent developments in a number of domains of the sciences. With several hundred machine learning (ML) algorithms and models, and numerous AI-specific hardware platforms, a common quest for all scientists working on AI for Science is around the selection of machine learning algorithm(s) to solve their domain-specific scientific problems. A number of different initiatives around AI Benchmarking have been set up and have been useful in understanding the benefits of different ML algorithms for different tasks.However, with the majority of these AI Benchmarking initiatives focusing on the conventional notions of benchmarking, where the focus is purely runtime performance (such as training time or inference time), their suitability for benchmarking different ML algorithms for solving scientific problems has been viewed as a performance problem even though both are hardly the same. To make reasonable, explainable, and justifiable advancements in science using AI, it is critical to focus on the merits of these algorithms in handling different domain science problems. In other words, more emphasis must be given on Benchmarking for AI for Science than AI Benchmarking. The vision of the former is not only to assess the performance of ML algorithms, but also to assess, and understand the benefits and merits of different ML algorithms in handling scientific problems. Benchmarking for AI for Science, instead of pure performance focused AI Benchmarking, has several benefits: (i) it has the potential to offer advances in the sciences, much more rapidly than through pure performance-based AI methods, (ii) it will encourage the community to focus on developing better domain-specific AI techniques, particularly given the provision for being able to benchmark different techniques, and (iii) it will encourage hardware manufacturers to focus on developing science-specific hardware subsystems.

Thiyagalingam, Jeyan↗

DOE OSTI2025

The Effects of Compounded Model Size Reductions on Adversarial Robustness

Recent advances in Edge AI and Tiny Machine Learning (TinyML) have enabled the deployment of machine learning models on resource-constrained environments. However, deploying these models on edge devices, such as micro-controllers, requires significant model footprint reduction through a variety of techniques such as quantization, pruning, and clustering. While these optimization methods offer considerable advantages, they potentially introduce AI-related security vulnerabilities, particularly concerning model robustness with respect to adversarial AI attacks. Prior research has extensively examined the impact of quantization on adversarial robustness; however, the effects of alternative reduction techniques and their combinations remain understudied. This paper investigates the impact of model size reduction techniques on adversarial robustness, when applied individually and combined. We utilized Fast Gradient Sign Method (FGSM) and Projected Gradient Descent (PGD) attacks to generate adversarial perturbations for both training and testing data, and then evaluated the models' accuracy under adversarial training conditions. Our findings revealed that reduction techniques generally diminished robustness; although, combining techniques was not found to make robustness any worse than when applied individually. Moreover, specific techniques can potentially enhance resistance to small size perturbations. This research provides insights into the trade-offs between model size reduction and security, establishing a foundation for future investigations into improving adversarial training techniques and methodologies for maintaining robustness while preserving memory footprint benefits.

Austria, Phillipe [ORNL] (ORCID:0000000236223973)↗