Search NASA⌕ Search

SEARCH · Search NASA

Results for “model accuracy”

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 37 records · Page 2
All sourcesNASA NTRSDOE OSTI
DOE OSTI2021

Evaluate the impact of sensor accuracy on model performance in data-driven building fault detection and diagnostics using Monte Carlo simulation

The performance of data-driven fault detection and diagnostics (FDD) is heavily dependent on sensors. However, sensor inaccuracy and sensor faults are pervasive in building operation: inaccurate and missing sensor readings deteriorate FDD performance; sensor inaccuracy will also affect the selection of sensor for data-driven FDD in the model training process, which is another key factor of data-driven FDD performance. Sensor accuracy and sensor selection individually are well-studied research topics in this field, but the impact of sensor accuracy on sensor selection and its further impact on FDD performance has not been evaluated and quantified. In this paper, we developed a novel analysis methodology that comprehensively evaluates sensor fault on sensor selection and FDD accuracy. Monte Carlo simulation is applied to deal with multiple stochastic sensor inaccuracy and provide probabilistic analysis results of the impact of sensor inaccuracy on sensor selection and FDD accuracy. This methodology focuses on the net impact of fault states across a full sensor set. The developed methodology can be used for the early-stage sensor design and operation-stage sensor maintenance. Furthermore, a case study is conducted to demonstrate the analysis methodology using a commercial building model crated to Flexible Research Platform located at Oak Ridge National Laboratory, USA.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗
DOE OSTI2023

AI-accelerated protein-ligand docking for SARS-CoV-2 is 100-fold faster with no significant change in detection

Protein-ligand docking is a computational method for identifying drug leads. The method is capable of narrowing a vast library of compounds down to a tractable size for downstream simulation or experimental testing and is widely used in drug discovery. While there has been progress in accelerating scoring of compounds with artificial intelligence, few works have bridged these successes back to the virtual screening community in terms of utility and forward-looking development. We demonstrate the power of high-speed ML models by scoring 1 billion molecules in under a day (50 k predictions per GPU seconds). We showcase a workflow for docking utilizing surrogate AI-based models as a pre-filter to a standard docking workflow. Our workflow is ten times faster at screening a library of compounds than the standard technique, with an error rate less than 0.01% of detecting the underlying best scoring 0.1% of compounds. Our analysis of the speedup explains that another order of magnitude speedup must come from model accuracy rather than computing speed. In order to drive another order of magnitude of acceleration, we share a benchmark dataset consisting of 200 million 3D complex structures and 2D structure scores across a consistent set of 13 million “in-stock” molecules over 15 receptors, or binding sites, across the SARS-CoV-2 proteome. We believe this is strong evidence for the community to begin focusing on improving the accuracy of surrogate models to improve the ability to screen massive compound libraries 100 × or even 1000 × faster than current techniques and reduce missing top hits. The technique outlined aims to be a fast drop-in replacement for docking for screening billion-scale molecular libraries.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2024

Spike-and-Slab Shrinkage Priors for Structurally Sparse Bayesian Neural Networks

Network complexity and computational efficiency have become increasingly significant aspects of deep learning. Sparse deep learning addresses these challenges by recovering a sparse representation of the underlying target function by reducing heavily overparameterized deep neural networks. Specifically, deep neural architectures compressed via structured sparsity (e.g., node sparsity) provide low-latency inference, higher data throughput, and reduced energy consumption. In this article, we explore two well-established shrinkage techniques, Lasso and Horseshoe, for model compression in Bayesian neural networks (BNNs). To this end, we propose structurally sparse BNNs, which systematically prune excessive nodes with the following: 1) spike-and-slab group Lasso (SS-GL) and 2) SS group Horseshoe (SS-GHS) priors, and develop computationally tractable variational inference, including continuous relaxation of Bernoulli variables. We establish the contraction rates of the variational posterior of our proposed models as a function of the network topology, layerwise node cardinalities, and bounds on the network weights. Furthermore, we empirically demonstrate the competitive performance of our models compared with the baseline models in prediction accuracy, model compression, and inference latency.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2024

Progressive transfer learning for advancing machine learning-based reduced-order modeling

Abstract To maximize knowledge transfer and improve the data requirement for data-driven machine learning (ML) modeling, a progressive transfer learning for reduced-order modeling (p-ROM) framework is proposed. A key concept of p-ROM is to selectively transfer knowledge from previously trained ML models and effectively develop a new ML model(s) for unseen tasks by optimizing information gates in hidden layers. The p-ROM framework is designed to work with any type of data-driven ROMs. For demonstration purposes, we evaluate the p-ROM with specific Barlow Twins ROMs (p-BT-ROMs) to highlight how progress learning can apply to multiple topological and physical problems with an emphasis on a small training set regime. The proposed p-BT-ROM framework has been tested using multiple examples, including transport, flow, and solid mechanics, to illustrate the importance of progressive knowledge transfer and its impact on model accuracy with reduced training samples. In both similar and different topologies, p-BT-ROM achieves improved model accuracy with much less training data. For instance, p-BT-ROM with four-parent (i.e., pre-trained models) outperforms the no-parent counterpart trained on data nine times larger. The p-ROM framework is poised to significantly enhance the capabilities of ML-based ROM approaches for scientific and engineering applications by mitigating data scarcity through progressively transferring knowledge.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2021

DeepMerge – II. Building robust deep learning algorithms for merging galaxy identification across domains

In astronomy, neural networks are often trained on simulation data with the prospect of being used on telescope observations. Unfortunately, training a model on simulation data and then applying it to instrument data leads to a substantial and potentially even detrimental decrease in model accuracy on the new target dataset. Simulated and instrument data represent different data domains, and for an algorithm to work in both, domain-invariant learning is necessary. Here we employ domain adaptation techniques— Maximum Mean Discrepancy (MMD) as an additional transfer loss and Domain Adversarial Neural Networks (DANNs)— and demonstrate their viability to extract domain-invariant features within the astronomical context of classifying merging and non-merging galaxies. Additionally, we explore the use of Fisher loss and entropy minimization to enforce better in-domain class discriminability. We show that the addition of each domain adaptation technique improves the performance of a classifier when compared to conventional deep learning algorithms. We demonstrate this on two examples: between two Illustris-1 simulated datasets of distant merging galaxies, and between Illustris-1 simulated data of nearby merging galaxies and observed data from the Sloan Digital Sky Survey. The use of domain adaptation techniques in our experiments leads to an increase of target domain classification accuracy of up to ~20%. With further development, these techniques will allow astronomers to successfully implement neural network models trained on simulation data to efficiently detect and study astrophysical objects in current and future large-scale astronomical surveys.

galaxies: interactions↗
DOE OSTI2021

Fast convolutional neural networks on FPGAs with hls4ml

We introduce an automated tool for deploying ultra low-latency, low-power deep neural networks with convolutional layers on field-programmable gate arrays (FPGAs). By extending the hls4ml library, we demonstrate an inference latency of 5 µs using convolutional architectures, targeting microsecond latency applications like those at the CERN Large Hadron Collider. Considering benchmark models trained on the Street View House Numbers Dataset, we demonstrate various methods for model compression in order to fit the computational constraints of a typical FPGA device used in trigger and data acquisition systems of particle detectors. In particular, we discuss pruning and quantization-aware training, and demonstrate how resource utilization can be significantly reduced with little to no loss in model accuracy. We show that the FPGA critical resource consumption can be reduced by 97% with zero loss in model accuracy, and by 99% when tolerating a 6% accuracy degradation.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗
DOE OSTI2025

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]↗
DOE OSTI2023

Data-driven modeling to enhance municipal water demand estimates in response to dynamic climate conditions

Altered precipitation and temperature patterns from a changing climate will affect supply, demand, and overall municipal water system operations throughout the arid western U.S. While supply forecasts leverage hydrological models to connect climate influences with surface water availability, demand forecasts typically estimate water use independent of climate and other externalities. Stemming from an increased focus on seasonal water demand management, we use the Salt Lake City, Utah municipal water system as a test bed to assess model accuracy versus complexity trade-offs between simple climate-independent econometric-based models and complex climate-sensitive data-driven models to average to extreme wet and dry climate conditions—representative of a new climate normal. Here, the climate-independent model displayed low performance during extreme dry conditions with predictions exceeding 90% and 40% of the observed monthly and seasonal volumetric demands, respectively, which we attribute to insufficient model complexity. The climate-sensitive models displayed greater accuracy in all conditions, with an ordinary least squares model demonstrating a measurable reduction in prediction bias (3.4% vs. -27.3%) and RMSE (74.0 lpcd vs. 294 lpcd) compared to the climate-independent model. The climate-sensitive workflow increased model accuracy and characterized climate-demand interactions, demonstrating a novel tool to enhance water system management.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2023

Development of a Machine-Learned Cruise Guide Indicator for Rotorcraft

This paper presents a machine-learned virtual cruise guide indicator (vCGI) for Chinook helicopters. Two temporal neural networks were trained and evaluated on measured data from 55 flight tests, one for the fore rotor and another for the aft rotor, to predict a vCGI value, which protects 23 components from fatigue damage during steady-state conditions. Three different classes of machine learning architectures were evaluated for prediction of the vCGI from time sequences: a temporal convolutional neural network with 1D dilated causal convolutions, a long short-term memory recurrent neural network, and an attention-based transformer architecture. The final average model accuracy on unseen flight data is currently greater than 93% for CGI values which could result in fatigue damage and 90% for normal operation CGI values. Model accuracy was improved through a series of advancements in:(1) selection of optimal training data using temporal collective variables and unsupervised learning, (2) dataset augmentation with maximum-entropy temporal collective variables, and (3) implementation of a mixture-of-experts classification- regression approach using an adversarial classification approach to assign maneuver labels. The results are presented for each advancement in model development along with lessons learned in training machine learning models on real- world, time-dependent rotorcraft data.

Boyer, Mathew↗
DOE OSTI2025

FedCSpc: A Cross-Silo Federated Learning System With Error-Bounded Lossy Parameter Compression

Cross-Silo federated learning is widely used for scaling deep neural network (DNN) training over data silos from different locations worldwide while guaranteeing data privacy. Communication has been identified as the main bottleneck when training large-scale models due to large-volume model parameters and gradient transmission across public networks with limited bandwidth. Most previous works focus on gradient compression, while limited work tries to compress parameters that can not be ignored and extremely affect communication performance during the training. Here, to bridge this gap, we propose FedCSpc: an efficient cross-silo federated learning system with an XAI-driven adaptive parameter compression strategy for large-scale model training. Our work substantially differs from existing gradient compression techniques due to the distinct data features of gradient and parameter. The key contributions of this paper are fourfold. (1) Our designed FedCSpc proposes to compress the parameter during the training using the state-of-the-art error-bounded lossy compressor – SZ3. (2) We develop an adaptive compression error bound adjustment algorithm to guarantee the model accuracy effectively. (3) We exploit an efficient approach to utilize the idle CPU resources of clients to compress the parameters. (4) We perform a comprehensive evaluation with a wide range of models and benchmarks on a GPU cluster with 65 GPUs. Results show that FedCSpc can achieve the same model accuracy as FedAvg while reducing the data volume of parameters and gradients in communication by up to 7.39× and 288×, respectively. With 32 clients on a 4 Gb size model, FedCSpc significantly outperforms FedAvg in wall-clock time in the emulated WAN environment (at the bandwidth of 1 Gbps or lower without loss of generality).

SZ3↗
DOE OSTI2021

Predicting nepheline precipitation in waste glasses using ternary submixture model and machine learning

Nepheline precipitation in nuclear waste glasses during vitrification can be detrimental due to its negative effect on chemical durability. Developing models to accurately predict nepheline precipitation from compositions is important to increase waste loading since existing models can be overly conservative. In this study, an expanded dataset containing 955 glasses was compiled from literature data, where 355 glasses are for high-level waste (HLW). Previously developed submixture models were refitted using the new dataset, where a misclassification rate of 7.8% was achieved. Nine machine learning (ML) algorithms (e.g., k-nearest neighbor, Gaussian process regression, artificial neural network, support vector machine, decision tree, etc.) were applied to evaluate their ability of predicting nepheline precipitation from compositions. Model accuracy, precision, recall/sensitivity, and F1 score were systemically compared between different ML algorithms and modeling protocols. Good model prediction with an accuracy ~0.9 (misclassification rate of ~10%) was observed with different algorithms under certain protocol. This study evaluated various ML models to predict nepheline precipitations in waste glasses, highlighting the importance of data preparation, modeling protocol, and their effect on model stability and reproducibility. The results provide insights into applying ML to predict glass properties and suggest areas for future research on modeling nepheline precipitations.

Lu, Xiaonan↗
DOE OSTI2026

Simulation of Channel Flow with Square Ribs for Blanket First-Wall Cooling: Geometry-Specific Tuning of k-ω Model Using Adjoint Method

Cooling of the plasma-facing first wall is challenging in the design of blanket components because of the high heat flux (on the order of 𝑀𝑊/𝑚2) from the plasma, especially when a low thermal mass medium like helium is chosen as the coolant. Therefore, heat transfer enhancement in which the convective heat transfer rate is augmented by the addition of turbulence-promoting structures becomes a key initiative for providing sufficient cooling capability with helium. Previously, computational fluid dynamics simulations had been performed on pipe flows with different transverse and longitudinal ribbed geometries at Oak Ridge National Laboratory to compare the enhancement performance among different ribbed geometries. Rib shape morphing had been conducted to obtain an optimized rib profile. In the work presented here, the adjoint method is adopted in the ANSYS Fluent solver for turbulence model augmentation, and the Generalized k-ω (GEKO) turbulence model is employed because of its ability of tuning the turbulence model. The Nusselt number and pressure drop obtained from the channel flow with bottom ribbed wall experiments are used as the targets. Sensitivity analysis provides information as guidance to improve the turbulence model accuracy. The augmented GEKO model is tuned for the studied ribbed channel geometry and flow conditions, providing improved predictive accuracy within this context. Extension to other configurations offers potential but may require additional tuning and validation.

Xu, Tracy [ORNL] (ORCID:0009000193700887)↗
DOE OSTI2023

A Method for Projecting Cloud Shadows Onto a Central Receiver Field to Predict Receiver Damage

This work demonstrates methods of mapping high-spatial-resolution direct normal irradiance (DNI) data from satellites, Total Sky Imagers (TSIs), and analogous data sources onto a heliostat field for characterizing the spatial and temporal variation of the incident flux on a central receiver tower during cloud transient events. The mapping methods are incorporated into an optical software module that interfaces with CoPylot–SolarPILOT’s python API– to provide computationally efficient optical simulation of the heliostat field and the solar power tower. Eventually, this optical model will be incorporated into optimization models whereby a plant operator can understand the effects of cloud transient events on overall power production and receiver lifetime due to creep-fatigue damage and therefore make better informed decisions about receiver shutdown events. By more accurately modelling the effects of cloud events on receiver flux maps, this work may determine the magnitude and frequency of thermal cycling on receiver tubes and panels using actual or realistic cloud shapes instead of averaged DNI values–which may undercount the total cycle number. This work may also prevent unnecessary plant shutdowns due to overly precautionary control strategies and characterize the relative impact of various cloud types on receiver life. We plan to eventually integrate this methodology into the System Advisor Model (SAM) to improve performance model accuracy during periods of cloudiness. In this paper, we demonstrate generating DNI maps and mapping them to a solar field in CoPylot using 10 m resolution data from publicly available Sentinel-2 satellite data over the Crescent Dunes plant.

Mullin, Matthew↗
DOE OSTI2023

Improving E3SM Land Model Photosynthesis Parameterization via Satellite SIF, Machine Learning, and Surrogate Modeling

The parameterization of key photosynthesis parameters is one of the key uncertain sources in modeling ecosystem gross primary productivity (GPP). Solar-induced chlorophyll fluorescence (SIF) offers a good proxy for GPP since it marks the actual process of photosynthesis; while machine learning (ML) provides a robust approach to model the GPP-SIF relationship. Here, we trained the boosted regressing tree (BRT) and the Random Forest ML models with Greenhouse Gases Observing Satellite SIF data and in situ GPP observations from 49 eddy covariance towers. These trained ML GPP-SIF models were fed into the Energy Exascale Earth System Model (E3SM) Land Model (ELM) to generate ELM-simulated global SIF estimates, which were then benchmarked against satellite SIF observations with a surrogate modeling approach. Our results indicated good modeling performance of the ML-based GPP-SIF relationship. The ELM model when fed with the ML GPP-SIF models also can well predict the spatial-temporal variations in SIF. We also found high model accuracy for the surrogate modeling. Model parameter sensitivity analysis suggested that the fraction of leaf nitrogen in RuBisCO (flnr) is the most sensitive parameter to the SIF; other sensitive parameters include the Ball-Berry stomatal conductance slope (mbbopt) and the vcmax entropy (vcmaxse). The posterior uncertainty in simulated GPP was greatly reduced after benchmarking, and the model produced improved spatial patterns of mean GPP relative to FLUXCOM GPP. Our integrated approach provides a new avenue for improving land models and using remote-sensing SIF, which can be further improved in the future with more ground- and satellite-based observations.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2023

Quantum machine learning with differential privacy

Abstract Quantum machine learning (QML) can complement the growing trend of using learned models for a myriad of classification tasks, from image recognition to natural speech processing. There exists the potential for a quantum advantage due to the intractability of quantum operations on a classical computer. Many datasets used in machine learning are crowd sourced or contain some private information, but to the best of our knowledge, no current QML models are equipped with privacy-preserving features. This raises concerns as it is paramount that models do not expose sensitive information. Thus, privacy-preserving algorithms need to be implemented with QML. One solution is to make the machine learning algorithm differentially private, meaning the effect of a single data point on the training dataset is minimized. Differentially private machine learning models have been investigated, but differential privacy has not been thoroughly studied in the context of QML. In this study, we develop a hybrid quantum-classical model that is trained to preserve privacy using differentially private optimization algorithm. This marks the first proof-of-principle demonstration of privacy-preserving QML. The experiments demonstrate that differentially private QML can protect user-sensitive information without signficiantly diminishing model accuracy. Although the quantum model is simulated and tested on a classical computer, it demonstrates potential to be efficiently implemented on near-term quantum devices [noisy intermediate-scale quantum (NISQ)]. The approach’s success is illustrated via the classification of spatially classed two-dimensional datasets and a binary MNIST classification. This implementation of privacy-preserving QML will ensure confidentiality and accurate learning on NISQ technology.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2023

A leaf-level spectral library to support high-throughput plant phenotyping: predictive accuracy and model transfer

Abstract Leaf-level hyperspectral reflectance has become an effective tool for high-throughput phenotyping of plant leaf traits due to its rapid, low-cost, multi-sensing, and non-destructive nature. However, collecting samples for model calibration can still be expensive, and models show poor transferability among different datasets. This study had three specific objectives: first, to assemble a large library of leaf hyperspectral data (n=2460) from maize and sorghum; second, to evaluate two machine-learning approaches to estimate nine leaf properties (chlorophyll, thickness, water content, nitrogen, phosphorus, potassium, calcium, magnesium, and sulfur); and third, to investigate the usefulness of this spectral library for predicting external datasets (n=445) including soybean and camelina using extra-weighted spiking. Internal cross-validation showed satisfactory performance of the spectral library to estimate all nine traits (mean R2=0.688), with partial least-squares regression outperforming deep neural network models. Models calibrated solely using the spectral library showed degraded performance on external datasets (mean R2=0.159 for camelina, 0.337 for soybean). Models improved significantly when a small portion of external samples (n=20) was added to the library via extra-weighted spiking (mean R2=0.574 for camelina, 0.536 for soybean). The leaf-level spectral library greatly benefits plant physiological and biochemical phenotyping, whilst extra-weight spiking improves model transferability and extends its utility.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2020

Parallel Embeddings: a Visualization Technique for Contrasting Learned Representations

We introduce ``Parallel Embeddings'', a new technique that generalizes the classical Parallel Coordinates visualization technique from a sequence of features within a single matrix to a sequence of multiple feature matrices. This is accomplished by separately embedding each matrix in the sequence into a 2-D frame, juxtaposing these 2-D representations, then connecting points between frames. We outline how this visualization can be used for several ``model comparison'' tasks in machine learning. We compare user performance with Parallel Embeddings to a baseline using the TensorFlow Embedding Projector for estimating model accuracy and understanding conceptual model differences. We found that users were more accurate with Parallel Embeddings for model comparison and that users learned how to use Parallel Embeddings more quickly than TensorFlow Embedding Projector. Furthermore, users' analytical process using Embedding Projector was positively affected by using Parallel Embeddings beforehand.

Model comparison, machine learing, machine learnin↗