Search NASA⌕ Search

SEARCH · Search NASA

Results for “Distributed training”

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 343 records · Page 19

Multi-Sensor Data from A-Train Instruments Brought Together for Atmospheric Research

The A-Train is comprised of a series of instruments, developed independently, that measure highly related atmospheric components along the same flight path. In order to intercompare data from this multitude of sensors, researchers must access, subset, visualize, analyze and correlate distributed atmosphere measurements from the various A-Train instruments. The A-Train Data Depot (ATDD) has been operational for over a year, successfully performing the aforementioned functions on behalf of researchers, thus providing co-registered data from the Cloudsat, CALIOP, AIRS, and MODIS instruments for further intercomparisons. Of late, significant data from OM1 and POLDER are now included in the 'depot'. By specifying the desired spatial and temporal range, the researcher can subset, visualize, co-register, and access multi-sensor A-Train data related to: Cloud, aerosol, atmospheric temperature, and water vapor parameters (vertical profile visualizations); Cloud Pressure, cloud top temperature, water vapor, cloud optical thickness, and aerosol products (horizontal strips subsetted +/- 100km from the profile visualizations), and; Cloud pressure parameters (2-D line plots overlayed on the vertical profiles). All data is plotted using the GIOVANNI data exploration tool. A new feature of GIOVANNI is its ability to have collocated and subsetted data sets as well as PNG image files downloaded to the researcher's computing facility. By providing a convenient way to visualize and acquire multi-sensor data, ATDD affords users more time and effort to further their research.

Smith, Peter M.↗

Cations Control Lipid Bilayer Memcapacitance Associated with Long-Term Potentiation

Phospholipid bilayers can be described as capacitors whose capacitance per unit area (specific capacitance, C m ) is determined by their thickness and dielectric constant–independent of applied voltage. It is also widely assumed that the C m of membranes can be treated as a “biological constant”. Recently, using droplet interface bilayers (DIBs), it was shown that zwitterionic phosphatidylcholine (PC) lipid bilayers can act as voltage-dependent, nonlinear memory capacitors, or memcapacitors. When exposed to an electrical “training” stimulation protocol, capacitive energy storage in lipid membranes was enhanced in the form of long-term potentiation (LTP), which enables biological learning and long-term memory. LTP was the result of membrane restructuring and the progressive asymmetric distribution of ions across the lipid bilayer during training, which is analogous, for example, to exponential capacitive energy harvesting from self-powered nanogenerators. Here, we describe how LTP could be produced from a membrane that is continuously pumped into a nonequilibrium steady state, altering its dielectric properties. During this time, the membrane undergoes static and dynamic changes that are fed back to the system’s potential energy, ultimately resulting in a membrane whose modified molecular structure supports long-term memory storage and LTP. Here, we also show that LTP is very sensitive to different salts (KCl, NaCl, LiCl, and TmCl 3 ), with LiCl and TmCl 3 having the most profound effect in depressing LTP, relative to KCl. This effect is related to how the different cations interact with the bilayer zwitterionic PC lipid headgroups primarily through electric-field-induced changes to the statistically averaged orientations of water dipoles at the bilayer headgroup interface.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Li-ion Battery Material phase prediction through Hierarchical Curriculum Learning

Li-ion Batteries (LIB), one of the most efficient energy storage devices, are widely adopted in many industrial applications. Imaging data of these battery electrodes obtained from X-ray tomography can explain the distribution of material constituents and allow reconstructions to study electron transport pathways. Therefore, it can eventually help quantify various associated properties of electrodes (e.g., volume-specific surface area, porosity) which determine the performance of batteries. However, these images often suffer from low image contrast between multiple material constituents , making it difficult for humans to distinguish and characterize these constituents through visualization. A minor error in detecting distributions among the material constituents can lead to a high error in the calculated parameters of material properties.We present a novel hierarchical curriculum learning framework to address the complex task of estimating material constituent distribution in battery electrodes. To provide spatially smooth prediction, our framework comprises three modules: (i) an uncertainty-aware model trained to yield inferences conditioned upon global knowledge of material distribution, (ii) a technique to capture relatively more fine-grained (local) distributional signals, (iii) an aggregator to appropriately fuse the local and global effects towards obtaining the final distribution.

Tabassum, Anika↗

Detecting gravitational waves from inspiraling binaries with a network of geographically separated detectors: Coherent versus coincident strategies

We compare two strategies of multidetector detection of compact binary inspiral signals, namely, the coincidence and the coherent for the realistic case of geographically separated detectors. The naive coincident strategy treats the detectors as if they are isolated - compares individual detector statistics with their respective thresholds while the coherent strategy combines the detector network data coherently to obtain a single detection statistic which is then compared with a single threshold. We also consider an enhanced coincidence strategy which is intermediate in the sense that though the individual statistics are added in quadrature and the sum compared with a single threshold, the estimated parameters are also checked for consistency. For simplicity, we consider detector pairs having the same power spectral density of noise, as that of initial LIGO and also assume the noise to be stationary and Gaussian. Further, since we consider the detectors to be widely separated on Earth, we take the instrumental noises to be uncorrelated; the wide separation implicitly means that since the detector arms must lie parallel to the Earth's surface, the detectors necessarily have different orientations. We compare the performances of the methods by plotting the receiver operating characteristics for the strategies. Several results are derived analytically in order to gain insight. Simulations are performed in order to plot the receiver operating characteristic (ROC) curves. A single astrophysical source as well as a distribution of sources is considered. We assume a 1 yr data train and a mass range of 1-40M{sub {center_dot}} for the case of astrophysically distributed sources. We find that the coherent strategy is superior to the two coincident strategies that we consider. Remarkably, the detection probability of the coherent strategy is 50% better than the naive coincident strategy. One the other hand, the difference in performance between the coherent strategy and enhanced coincident strategy is not very large. Even in this situation, it is not difficult to perform the real data analysis with the coherent strategy. The bottom line is that the coherent strategy is a good detection strategy.

99 GENERAL AND MISCELLANEOUS↗

Normalizing flows for domain adaptation when identifying Λ hyperon events

Here this study focuses on the application of a normalizing flow as a method of domain adaptation when classifying physics data. Normalizing flows offer a way to transform data points between two different distributions. The present study investigates a novel method of transforming latent representations of physics data to a normal distribution and then to a physics distribution again. The final distribution models a simulated distribution. After being transformed, the data can be classified by a neural network trained on labeled simulation data. The present study succeeds in training two normalizing flows that can transform between data (or simulation) and a Gaussian distribution.

47 OTHER INSTRUMENTATION↗

EV Charging Infrastructure Energization An Overview of Approaches for Simplifying and Accelerating Timelines to Processing EV Charging Load Service Requests

The United States has seen significant growth in electric vehicle (EV) adoption, leading to increased demand for EV charging infrastructure. Over the past decade, EV charging infrastructure site developers, site hosts, and electric distribution utilities have navigated the process to integrate chargers onto the electric grid. Site developers and site hosts have raised the alarm that the integration process for high-powered EV charging projects does not meet the needs of the EV market for timeliness or cost. High-powered charging stations typically require a load service request or an agreement with the local utility to connect to the grid. The process of energizing a new high-powered charging site can be complex and time-consuming, often taking up to 2 years. This timeline is the result of current utility energization processes having been designed for construction projects that take longer to build (i.e., buildings). The specific challenges stem from various factors, including compartmentalization in application processes, the integration of EV charging process approvals with other distributed energy resources (DERs), and the need to ensure grid reliability. The energization process needs to evolve to meet the growing demand for high-powered EV charging. This white paper compiles information gathered through various conversations with key stakeholders, including utilities, utility regulators, EV charging operators, site developers, and authorities having jurisdiction (AHJ) as well as through an extensive literature review. This document identifies the challenges and provides potential solutions to streamline the process of connecting EV charging infrastructure to the power grid in the United States, serving as a starting point for future conversations around these solutions. The solutions noted in this white paper require collaborative efforts among utilities, regulators, and EV charging infrastructure developers to streamline the grid connection process for EV charging infrastructure. They are broadly organized into four areas: 1. Increase data access and transparency: Develop automated load service request tools, integrate hosting capacity and load service request analyses, incorporate EV adoption forecasts, and provide transparency on the processing queue. 2. Improve energization processes and timing: Create fast-track options based on prescreening criteria, provide flexibility or phased approvals in the load service request/interconnection process, build internal knowledge within utilities about EV charging technologies, and provide standardized workforce training. 3. Promote economic efficiency: Right size distribution components to accurately reflect the load requirements of EV charging infrastructure, make proactive investments in grid infrastructure based on EV adoption forecasts and growth projections, and consider energy equity and environmental justice factors such as equitable access to EV charging when planning infrastructure. 4. Improve grid reliability and resilience: Use load management/power control systems (PCS) at EV charging stations, adopt and implement harmonized standards for communication protocols and information models between the EV charging and grid control infrastructure, and address cybersecurity considerations by implementing robust security measures and standards for EV charging infrastructure—with particular emphasis on clarifying the security requirements for the interface to the grid. The objective of the solutions proposed in this white paper is to accelerate the timeline and decrease costs associated with connecting EV charging infrastructure to the grid. Electric utilities, utility regulators, EV charging infrastructure developers, and site hosts will first need to understand which solutions are available in their service territory, and if warranted, which combination of solutions would support their specific needs. Through the successful implementations of solutions at scale detailed here, industry will demonstrate a new and innovative ecosystem where timely deployment and energization of EV charging infrastructure with greater grid resiliency and reliability is a reality.

24 POWER TRANSMISSION AND DISTRIBUTION↗

On the Generalizability of Time-of-Flight Convolutional Neural Networks for Noninvasive Acoustic Measurements

Bulk wave acoustic time-of-flight (ToF) measurements in pipes and closed containers can be hindered by guided waves with similar arrival times propagating in the container wall, especially when a low excitation frequency is used to mitigate sound attenuation from the material. Convolutional neural networks (CNNs) have emerged as a new paradigm for obtaining accurate ToF in non-destructive evaluation (NDE) and have been demonstrated for such complicated conditions. However, the generalizability of ToF-CNNs has not been investigated. In this work, we analyze the generalizability of the ToF-CNN for broader applications, given limited training data. We first investigate the CNN performance with respect to training dataset size and different training data and test data parameters (container dimensions and material properties). Furthermore, we perform a series of tests to understand the distribution of data parameters that need to be incorporated in training for enhanced model generalizability. This is investigated by training the model on a set of small- and large-container datasets regardless of the test data. We observe that the quantity of data partitioned for training must be of a good representation of the entire sets and sufficient to span through the input space. The result of the network also shows that the learning model with the training data on small containers delivers a sufficiently stable result on different feature interactions compared to the learning model with the training data on large containers. To check the robustness of the model, we tested the trained model to predict the ToF of different sound speed mediums, which shows excellent accuracy. Furthermore, to mimic real experimental scenarios, data are augmented by adding noise. We envision that the proposed approach will extend the applications of CNNs for ToF prediction in a broader range.

47 OTHER INSTRUMENTATION↗

Novel tools for stepping source brachytherapy treatment planning: Enhanced geometrical optimization and interactive inverse planning

Purpose: Dose optimization for stepping source brachytherapy can nowadays be performed using automated inverse algorithms. Although much quicker than graphical optimization, an experienced treatment planner is required for both methods. With automated inverse algorithms, the procedure to achieve the desired dose distribution is often based on trial-and-error. Methods: A new approach for stepping source prostate brachytherapy treatment planning was developed as a quick and user-friendly alternative. This approach consists of the combined use of two novel tools: Enhanced geometrical optimization (EGO) and interactive inverse planning (IIP). EGO is an extended version of the common geometrical optimization method and is applied to create a dose distribution as homogeneous as possible. With the second tool, IIP, this dose distribution is tailored to a specific patient anatomy by interactively changing the highest and lowest dose on the contours. Results: The combined use of EGO–IIP was evaluated on 24 prostate cancer patients, by having an inexperienced user create treatment plans, compliant to clinical dose objectives. This user was able to create dose plans of 24 patients in an average time of 4.4 min/patient. An experienced treatment planner without extensive training in EGO–IIP also created 24 plans. The resulting dose-volume histogram parameters were comparable to the clinical plans and showed high conformance to clinical standards. Conclusions: Even for an inexperienced user, treatment planning with EGO–IIP for stepping source prostate brachytherapy is feasible as an alternative to current optimization algorithms, offering speed, simplicity for the user, and local control of the dose levels.

60 APPLIED LIFE SCIENCES↗

Joint aircraft loading/structure response statistics of time to service crack initiation

A reliability analysis for predicting the statistical distribution of time to fatigue crack initiation for aircraft structures in service is presented. The present analysis utilizes the statistical data of the specimen fatigue tests, the full-scale structure tests, and the statistical dispersion of aircraft service loads. The statistical distribution of the time to fatigue crack initiation of the full-scale structure under laboratory loading spectrum is assumed to be Weibull. The service loads for gust turbulences are modeled as Poisson processes for transport-type aircraft, while the maneuver loads are modeled as compound Poisson processes for fighter and training aircraft. It is found that the statistical distribution of time to fatigue crack initiation for aircraft structures in service is not Weibull and that the prediction on the basis of the Weibull distribution is unconservative, in particular in the early service time.

Yang, J.-N.↗

Operator learning for predicting multiscale bubble growth dynamics

We report simulating and predicting multiscale problems that couple multiple physics and dynamics across many orders of spatiotemporal scales is a great challenge that has not been investigated systematically by deep neural networks (DNNs). Herein, we develop a framework based on operator regression, the so-called deep operator network (DeepONet), with the long-term objective to simplify multiscale modeling by avoiding the fragile and time-consuming “hand-shaking” interface algorithms for stitching together heterogeneous descriptions of multiscale phenomena. To this end, as a first step, we investigate if a DeepONet can learn the dynamics of different scale regimes, one at the deterministic macroscale and the other at the stochastic microscale regime with inherent thermal fluctuations. Specifically, we test the effectiveness and accuracy of the DeepONet in predicting multirate bubble growth dynamics, which is described by a Rayleigh–Plesset (R–P) equation at the macroscale and modeled as a stochastic nucleation and cavitation process at the microscale by dissipative particle dynamics (DPD). First, we generate data using the R–P equation for multirate bubble growth dynamics caused by randomly time-varying liquid pressures drawn from Gaussian random fields (GRFs). Our results show that properly trained DeepONets can accurately predict the macroscale bubble growth dynamics and can outperform long short-term memory networks. We also demonstrate that the DeepONet can extrapolate accurately outside the input distribution using only very few new measurements. Subsequently, we train the DeepONet with DPD data corresponding to stochastic bubble growth dynamics. Although the DPD data are noisy and we only collect sparse data points on the trajectories, the trained DeepONet model is able to predict accurately the mean bubble dynamics for time-varying GRF pressures. Taken together, our findings demonstrate that DeepONets can be employed to unify the macroscale and microscale models of the multirate bubble growth problem, hence providing new insight into the role of operator regression via DNNs in tackling realistic multiscale problems and in simplifying modeling with heterogeneous descriptions.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

An Evaluation of Semiempirical Models for Partitioning Photosynthetically Active Radiation Into Diffuse and Direct Beam Components

Photosynthesis is more efficient under diffuse than direct beam photosynthetically active radiation (PAR) per unit PAR, but diffuse PAR is infrequently measured at research sites. We examine four commonly used semiempirical models (Erbs et al., 1982, https://doi.org/10.1016/0038-092X(82)90302-4; Gu et al., 1999, https://doi.org/10.1029/1999JD901068; Roderick, 1999, https://doi.org/10.1016/S0168- 1923(99)00028-3; Weiss & Norman, 1985, https://doi.org/10.1016/0168-1923(85)90020-6) that partition PAR into diffuse and direct beam components based on the negative relationship between atmospheric transparency and scattering of PAR. Radiation observations at 58 sites (140 site years) from the La Thuille FLUXNET data set were used for model validation and coefficient testing. All four models did a reasonable job of predicting the diffuse fraction of PAR (Φ) at the 30 min timescale, with site median r2 values ranging between 0.85 and 0.87, model efficiency coefficients (MECs) between 0.62 and 0.69, and regression slopes within 10% of unity. Model residuals were not strongly correlated with astronomical or standard meteorological variables. We conclude that the Roderick (1999, https://doi.org/10.1016/S0168- 1923(99)00028-3) and Gu et al. (1999, https://doi.org/10.1029/1999JD901068) models performed better overall than the two older models. Using the basic form of these models, the data set was used to find both individual site and universal model coefficients that optimized predictive accuracy. A new universal form of the model is presented in section 5 that increased site median MEC to 0.73. Site-specific model coefficients increased median MEC further to 0.78, indicating usefulness of local/regional training of coefficients to capture the local distributions of aerosols and cloud types.

54 ENVIRONMENTAL SCIENCES↗

The Role of National Research and Education Networks Providing Connectivity and Advanced Network Services to Virtual Communities in Collaborative R&E Projects. CUDI: The Mexican Case Part 1, , In: ISUM 2015: High Performance Computer Applications

This paper describes the role of National Research and Education Networks (NRENs) and their aim to provide a seamless, large scale, high-speed, low-latency network to accelerate scientific and education discovery for national, regional or global collaborative projects that are dependent on the use of modern high performance distributed computing systems (HPDCS) to process mathematical models, produce new knowledge and drive innovation when resolving complex problems. NRENs have become the mainstream of communication and collaboration to integrate the efforts of large geographically distributed teams of Virtual Communities (VC) that - train, teach and learn-, -create, use and share- knowledge, data, information and communications technology resources (e.g. computer and information systems, middleware, tools, instruments, data sets, repositories, applications and services) building e-infrastructures necessary to meet the collaborative projects’ goals. The paper focuses on the current status of the Mexican NREN: CUDI, its future organization and political support, necessary to provide seamless, large scale, high-speed, low latency and secure national backbone and international links providing connectivity, advanced network services and technical support to VCs and their e-infrastructures in collaborative R&E projects within the country, the LAC region and worldwide.

Computer Science↗

ThunderSecure: deploying real-time intrusion detection for 100G research networks by leveraging stream-based features and one-class classification network

Nowadays, data generated by large-scale scientific experiments are on the scale of petabytes per month. These data are transferred through dedicated high-bandwidth networks (40/100G) across distributed sites for processing, storage, and analysis. Like general purpose networks, research networks experience intrusions. However, monitoring anomalies in such high-speed network traffics is challenging given current cyber-infrastructure. Moreover, traditional network intrusion detection systems (NIDS) are signature based. However, anomaly patterns are difficult to define and that rulesets are often not updated frequently enough to reflect the changes of attack behaviors. We present ThunderSecure, a high-throughput, unsupervised learning-based intrusions detection system for 100G research networks. ThunderSecure implements an efficient packet processing and detection pipeline using multi-cores and GPUs. It extracts statistical and temporal features from real-time network data streams and feeds them to a one-class anomaly detection network. A baseline of normal distribution will be created based on the training observation. Testing traffic deviated from the learned profile will be marked as anomalies. We trained ThunderSecure on hundreds of billions of science data packets mirrored from two 100G network connections at Fermi National Accelerator Laboratory. The detection performance was evaluated on traffic captured from the same research network days and weeks after the training with different types of attack flows injected. Results show that ThunderSecure can recognize science data traffic captured long after the training and made nearly certain detection on the segment of the streams where anomalous flows were injected.

100G research network↗

SDYN-GANs: Adversarial learning methods for multistep generative models for general order stochastic dynamics

We introduce adversarial learning methods for data-driven generative modeling of dynamics of nth-order stochastic systems. Our approach builds on Generative Adversarial Networks (GANs) with generative model classes based on stable m-step stochastic numerical integrators. From observations of trajectory samples, we introduce methods for learning long-time predictors and stable representations of the dynamics. Our approaches use discriminators based on Maximum Mean Discrepancy (MMD), training protocols using both conditional and marginal distributions, and methods for learning dynamic responses over different time-scales. We show how our approaches can be used for modeling physical systems to learn force-laws, damping coefficients, and noise-related parameters. Our adversarial learning approaches provide methods for obtaining stable generative models for dynamic tasks including long-time prediction and developing simulations for stochastic systems.

• Artificial intelligence (AI) / machine learning ↗

Multi-head physics-informed neural networks for learning functional priors and uncertainty quantification

In numerous applications, the integration of prior knowledge and historical information is essential, particularly for tasks requiring the solution of ordinary or partial differential equations (ODEs/PDEs) in data-sparse or noisy environments. For instance, achieving accurate solutions to time-dependent PDEs with limited initial condition measurements necessitates an effective strategy for embedding prior knowledge. Hard-parameter sharing architectures in neural networks (NNs) have demonstrated success in both traditional and scientific machine learning domains, facilitating the learning of informative representations. Here, in this study, we introduce a novel, yet efficient, method to enhance physics-informed neural networks (PINNs) by incorporating a multi-head structure that enables the learning of functional priors from both empirical data and governing physical laws. This prior information can then be used to address data sparsity and high-level noise in solving ODE/PDE problems with uncertainty quantification (UQ). The approach, termed Multi-Head PINN (MH-PINN), consists of a shared body NN and multiple head NNs, each corresponding to an individual PINN instance. Our framework for functional prior learning is carried out in two stages: (1) training the MH-PINNs to develop a shared body NN alongside multiple head NNs, and (2) employing these trained head NNs to estimate a prior distribution through a normalizing flow-based density estimator. The learned functional prior can then be applied as a regularization mechanism in deterministic contexts or as an informative prior within a Bayesian inference framework, aiding in the resolution of subsequent ODE/PDE tasks. We evaluate the efficacy of MH-PINNs across five benchmark problems, including a high-dimensional parametric PDE, all characterized by data sparsity or substantial noise levels. Our findings reveal that MH-PINNs deliver accurate solutions and robust UQ, demonstrating adaptability across a range of complex and challenging scenarios.

Bayesian inference↗

Catlas: an automated framework for catalyst discovery demonstrated for direct syngas conversion

Catalyst discovery is paramount to support access to energy and key chemical feedstocks in a post fossil fuel era. Exhaustive computational searches of large material design spaces using ab initio methods like density functional theory (DFT) are infeasible. We seek to explore large design spaces at relatively low computational cost by leveraging large, generalized, graph-based machine learning (ML) models, which are pretrained and therefore require no upfront data collection or training. We present Catlas, a framework that distributes and automates the generation of adsorbate-surface configurations and ML inference of DFT energies to achieve this goal. Catlas is open source, making ML assisted catalyst screenings easy and available to all. To demonstrate its efficacy, we use Catlas to explore catalyst candidates for the direct conversion of syngas to multi-carbon oxygenates. For this case study, we explore 947 stable/metastable binary, transition metal intermetallics as possible catalyst candidates. On this subset of materials, we are able to predict the adsorption energy of key descriptors, *CO and *OH, with good accuracy (0.16, 0.14 eV MAE, respectively). Using the projected selectivity towards C 2+ oxygenates from an existing microkinetic model, we identified 144 candidate materials. For 10 promising candidates, DFT calculations reveal a good correlation with our assessment using ML. Finally, among the top elemental combinations were Pt–Ti, Pd–V, Ni–Nb, and Ti–Zn, all of which appear unexplored experimentally.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Optimisation of the Kaplan hydropower system via PID 2 and digital twin

Here, this paper proposes a proportional–integral-double–derivative (PID 2 ) optimisation method for the Kaplan hydropower system by building a digital twin. The study first uses one multilayer perceptron (MLP) to model the hydroturbine dynamic and then adopts three connected MLPs to model the generator dynamic, both in an open-loop fashion. Inspired by stochastic distribution control (SDC) theory, we regard the training of the turbine's neural network model as a process control problem, and we propose minimising entropy loss to update the network parameters. The next step is to build the digital twin by connecting the neural network models with a PID 2 controller and a lead-lag exciter and run the whole model in a closed-loop fashion. After that, a binary search approach is applied to optimise the PID 2 parameters based on the obtained digital twin model. The simulation results show that the proposed method can reduce the mean square tracking error by more than 90%. Furthermore, the method is extended to jointly optimise the PID 2 controller and excitation system gains through multiobjective optimisation, leveraging Pareto frontier analysis to balance active power and voltage tracking performance. Simulation results confirm the effectiveness of the proposed method, achieving a 83.46% reduction in relative mean square error of active power, a 47.13% reduction in terminal voltage tracking error, and an 82.78% improvement in the overall scalarized objective.

Hydropower system↗

Dynamical Sketching for Enhanced Communication Efficiency in Federated Learning

Federated learning (FL) has revolutionized distributed machine learning by enabling collaborative model training without sharing local data. However, communication efficiency and privacy guarantees remain significant challenges. This paper introduces a dynamic sketching mechanism in FL, optimizing the trade-off between communication efficiency and model accuracy. By dynamically selecting the sketch matrix size, our approach adapts to the evolving characteristics of the data and the model, ensuring optimal performance across diverse scenarios. We leverage Bayesian optimization to systematically tune the sketch parameters, achieving an effective balance between resource efficiency and model performance. Experimental results on the MNIST dataset using a convolutional neural network (CNN) architecture validate the proposed method's efficiency and scalability. Our dynamic sketching approach significantly outperforms fixed-size sketching techniques, achieving higher compression ratios (up to 62x) and providing better privacy guarantees while maintaining high model accuracy. These findings highlight the robustness and versatility of our approach and make it a valuable solution for privacy-preserving, communication-efficient federated learning.

Afrose, Sharmin [ORNL]↗