Search NASA⌕ Search

SEARCH · Search NASA

Results for “balance 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 127 records · Page 7
All sourcesNASA NTRSDOE OSTI
DOE OSTI2022

A Scalable Pipeline for Gigapixel Whole Slide Imaging Analysis on Leadership Class HPC Systems

Whole Slide Imaging (WSI) captures microscopic details of a patient's histopathological features at multiple res-olutions organized across different levels. Images produced by WSI are gigapixel-sized, and saving a single image in memory requires a few gigabytes which is scarce since a complicated model occupies tens of gigabytes. Performing a simple met-ric operation on these large images is also expensive. High-performance computing (HPC) can help us quickly analyze such large images using distributed training of complex deep learning models. One popular approach in analyzing these images is to divide a WSI image into smaller tiles (patches) and then train a simpler model with these reduced-sized but large numbers of patches. However, we need to solve three pre-processing challenges efficiently for pursuing this patch-based approach. 1) Creating small patches from a high-resolution image can result in a high number (hundreds of thousands per image) of patches. Storing and processing these images can be challenging due to a large number of I/O and arithmetic operations. To reduce I/Oand memory accesses, an optimal balance between the size and number of patches must exist to reduce I/O and memory accesses. 2) WSI images may have tiny annotated regions for cancer tissue and a significant portion with normal and fatty tissues; correct patch sampling should avoid dataset imbalance. 3) storing and retrieving many patches to and from disk storage might incur I/O latency while training a deep learning model. An efficient distributed data loader should reduce I/O latency during the training and inference steps. This paper explores these three challenges and provides empirical and algorithmic solutions deployed on the Summit supercomputer hosted at the Oak Ridge Leadership Computing Facility.

Dash, Sajal↗
DOE OSTI2023

Surface enrichment dictates block copolymer orientation

Orientation of block copolymer (BCP) morphology in thin films is critical to applications as nanostructured coatings. Despite being well-studied, the ability to control BCP orientation across all possible block constituents remains challenging. Here, in this study, we deploy coarse-grained molecular dynamics simulations to study diblock copolymer ordering in thin films, focusing on chain makeup, substrate surface energy, and surface tension disparity between the two constituent blocks. We explore the multi-dimensional parameter space of ordering using a machine-learning approach, where an autonomous loop using a Gaussian process (GP) control algorithm iteratively selects high-value simulations to compute. The GP kernel was engineered to capture known symmetries. The trained GP model serves as both a complete map of system response, and a robust means of extracting material knowledge. We demonstrate that the vertical orientation of BCP phases depends on several counter-balancing energetic contributions, including entropic and enthalpic material enrichment at interfaces, distortion of morphological objects through the film depth, and of course interfacial energies. BCP lamellae are found more resistant to these effects, and thus more robustly form vertical orientations across a broad range of conditions; while BCP cylinders are found to be highly sensitive to surface tension disparity.

77 NANOSCIENCE AND NANOTECHNOLOGY↗
DOE OSTI2025

Spectroscopic Online Monitoring: Using a Multi-Track Visible Spectrometer to Facilitate a Mass Balance Study in a Simulated TALSPEAK Process

Nuclear energy is a promising low-carbon energy candidate to meet the increased demand for green energy, where the integration of fuel recycling can have significant benefits for material usage and waste reduction. Utilizing in situ monitoring tools can provide ample opportunities to better control and safeguard nuclear material recycle processes while also offering knowledge and insight into real-time solution properties. The simultaneous measurement of analytical targets in multiple process locations can enable real-time mass balance and material accountancy calculations. This is demonstrated here with a mass balance study of Nd 3+ on countercurrent aqueous/organic metal extraction within a single centrifugal contactor. The Nd 3+ concentration was simultaneously monitored at the inlets and outlets of both aqueous and organic phases using a visible absorbance detector that allowed for the simultaneous measurement of up to six locations. The Nd 3+ concentration was calculated by using chemical data science algorithms, where model training sets were collected on a single track of the detector. The discussion includes addressing the challenges of using a model collected on a single track and applying it as a model across the other tracks on the detector. Each track of the detector corresponds to one measurement location on the contactor. The difference in the integrated moles of Nd 3+ between the inlet and outlet at the end of the experiment was near zero, indicating that the mass balance of this experiment was maintained. Overall, the online spectroscopic monitoring was able to follow changing solution conditions and accurately measure the concentration of Nd 3+ in different locations within the contactor system.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2022

Evaluation of satellite Leaf Area Index in California vineyards for improving water use estimation

Remote sensing estimation of evapotranspiration (ET) directly quantifies plant water consumption and provides essential information for irrigation scheduling, which is a pressing need for California vineyards as extreme droughts become more frequent. Many ET models take satellite-derived Leaf Area Index (LAI) as a major input, but how uncertainties of LAI estimations propagate to ET and the partitioning between evaporation and transpiration is poorly understood. Here we assessed six satellite-based LAI estimation approaches using Landsat and Sentinel-2 images against ground measurements from four vineyards in California and evaluated ET sensitivity to LAI in the thermal-based two-source energy balance (TSEB) model. We found that radiative transfer modeling-based approaches predicted low to medium LAI well, but they significantly underestimated high LAI in highly clumped vine canopies (RMSE ~ 0.97 to 1.27). Cubist regression models trained with ground LAI measurements from all vineyards achieved high accuracy (RMSE ~ 0.3 to 0.48), but these empirical models did not generalize well between sites. Red edge bands and the related vegetation index (VI) from the Sentinel-2 satellite contain complementary information of LAI to VIs based on near-infrared and red bands. TSEB ET was more sensitive to positive LAI biases than negative ones. Positive LAI errors of 50% resulted in up to 50% changes in ET, while negative biases of 50% in LAI caused less than 10% deviations in ET. However, even when ET changes were minimal, negative LAI errors of 50% led to up to a 40% reduction in modeled transpiration, as soil evaporation and plant transpiration responded to LAI change divergently. These findings call for careful consideration of satellite LAI uncertainties for ET modeling, especially for the partitioning of water loss between vine and soil or cover crop for effective vineyard irrigation management.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2025

Transfer learning nonlinear plasma dynamic transitions in low dimensional embeddings via deep neural networks

Deep learning algorithms provide a new paradigm to study high-dimensional dynamical behaviors, such as those in fusion plasma systems. Development of novel, data-driven model reduction methods, coupled with detection of abnormal modes with plasma physics, opens a unique opportunity to identify plasma instabilities through automated construction of parsimonious models that can be tuned to balance accuracy and cost. Our fusion transfer learning (FTL) model demonstrates success in rapidly reconstructing nonlinear kink mode structures by learning from a limited amount of nonlinear simulation data. The knowledge transfer process leverages a pre-trained neural encoder–decoder network, initially trained on linear simulations, to effectively capture nonlinear dynamics. The low-dimensional embeddings extract the coherent structures of interest, while preserving the inherent dynamics of the complex system. Experimental results highlight FTL’s capacity to capture transitional behaviors and dynamical features in plasma dynamics—a task often challenging for conventional methods. The model developed in this study is generalizable and can be extended broadly through transfer learning to address various magnetohydrodynamics modes.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2024

Reinforcement Learning for Distance Maximization in Energy-Constrained Robots

Our goal is to teach a robot to walk in an energy efficient way. Inspired by nature [1], we create a reinforcement learning method that balances distance traveled and energy consumption, which we call “utility.” Further, each step is symmetric to ensure equal strides while walking straight. We achieve our goal by rewarding decisions that lead to the best utility over several training iterations.

Pereira, Luiz Manella↗
DOE OSTI2020

14 Steps to Creating Value in Cost and Schedule Resources for Nuclear Waste Management - 20188

While Waste Management budgets are shrinking, DOE required cost and schedule controls are becoming increasingly stringent. At the same time, there is a scarcity of adept, qualified, and competent projects controls staff who both understand nuclear waste management and are capable of addressing today's cost and schedule challenges. In response, we have developed a systematic approach to training and mentoring project controls staff. The approach is based on the observations and lessons learned on projects, both successful and unsuccessful, over the past 20 years. During that time, we evaluated staff development and used that experience to design a program that progresses the maturity, knowledge and experience of the individual. The program is based on the unique needs of cost and schedule specialists while incorporating the best leadership and managerial applications. The system is not limited to developing individual software skills, which is a primary focus of many project controls training programs. Instead, our system develops an individual's soft and hard skill sets. The system develops the user's skill, talent, attitude, and drive to do the best job possible. It focuses on all aspects of becoming a more productive technical team while accepting responsibility for individual growth. It also enhances the project controls team by setting up checks and balances that everyone is aware of and participates in. The system is based on 14 steps of increasing complexity in knowledge and application. The system clarifies roles, responsibilities and expectations. It helps eliminate uncertainty and reduces stress on staff and managers. It contains everyday routine actions and is simple to implement. It allows the development of a consistent rhythm to staff growth and development. Mastering the steps will allow the user to achieve excellence in cost and schedule control. The 14 steps are: 1 - Key Performance Objectives, 2 - Check Lists, 3 - Desktop Guides, 4 - System Description, 5 - Procedures, 6 - Software and On the Job Training (OTJ), 7 - Traceability Checks, 8 - Surveillances, 9 - 32 Criteria, 10 - The Contract, 11 - Customer/Government Guidance, 12 - Training and Mentoring Others, 13 - Master your weekly and daily schedules, 14 - Improving your interpersonal skills. Application and mastery of this system will produce better cost and schedule control results. Your project controls staff will write better reports, analyze data better, make better use of their time, and be more valuable to decision makers. If followed rigorously, our system will produce a great project controls team, which is invaluable to the success of each project. (authors)

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗
DOE OSTI2026

Computational Modeling of Atmospheric Processes at Texas Southern University

Texas Southern University (TSU) is strengthening its research program in atmospheric chemistry and physics with a climate science emphasis by leveraging partnerships with the U.S. Department of Energy’s Atmospheric Radiation Measurement (ARM) Facility, Brookhaven National Laboratory (BNL), and the Tracking Aerosol Convection Interactions ExpeRiment (TRACER). This RDPP-supported program focuses on secondary organic aerosols (SOAs) and reactive atmospheric species that influence cloud formation, precipitation processes, and radiative forcing. SOAs play a critical role in cloud microphysics and Earth’s energy balance, yet the chemical and physical mechanisms governing SOA–cloud interactions remain a significant source of uncertainty in predictive climate models. Through computational modeling, observational data analysis, and national laboratory collaboration, this program develops a skilled cohort of students trained in atmospheric science, environmental data analysis, and climate-relevant modeling. These research experiences build technical competencies that are transferable to careers in government laboratories, academia, and industry. By engaging students from historically underrepresented communities in high-impact climate research, TSU expands participation in the atmospheric sciences workforce while contributing meaningful scientific insights to DOE-supported ARM research activities. This partnership strengthens national capacity in climate science and supports the development of the next generation of atmospheric researchers.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2023

Machine Learning Prediction of the Experimental Transition Temperature of Fe(II) Spin-Crossover Complexes

Spin-crossover (SCO) complexes are materials that exhibit changes in the spin state in response to external stimuli, with potential applications in molecular electronics. It is challenging to know a priori how to design ligands to achieve the delicate balance of entropic and enthalpic contributions needed to tailor a transition temperature close to room temperature. Here, we leverage the SCO complexes from the previously curated SCO-95 data set [Vennelakanti et al. J. Chem. Phys. 159, 024120 (2023)] to train three machine learning (ML) models for transition temperature (T 1/2 ) prediction using graph-based revised autocorrelations as features. We perform feature selection using random forest-ranked recursive feature addition (RF-RFA) to identify the features essential to model transferability. Of the ML models considered, the full feature set RF and recursive feature addition RF models perform best, achieving moderate correlation to experimental T 1/2 values. We then compare ML T 1/2 predictions to those from three previously identified best-performing density functional approximations (DFAs) which accurately predict SCO behavior across SCO-95, finding that the ML models predict T 1/2 more accurately than the best-performing DFAs. In addition, we study ML model predictions for a set of 18 SCO complexes for which only estimated T 1/2 values are available. Upon excluding outliers from this set, the RF-RFA RF model shows a strong correlation to estimated T 1/2 values with a Pearson’s r of 0.82. In contrast, DFA-predicted T 1/2 values have large errors and show no correlation to estimated T 1/2 values over the same set of complexes. Overall, our study demonstrates slightly superior performance of ML models in comparison with some of the best-performing DFAs, and we expect ML models to improve further as larger data sets of SCO complexes are curated and become available for model training.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2025

Artificial neural networks estimate evapotranspiration for Miscanthus × giganteus as effectively as empirical model but with fewer inputs

Estimating actual evapotranspiration (ET) is particularly crucial for addressing how vegetation affects the water balance of ecosystems. ET estimation can be complex with empirical models due to their many parameters and reliance on aridity. In contrast, artificial neural networks (ANNs) could potentially estimate ET with fewer and more common meteorological parameters. In this study, we trained two ANNs, one using a feed-forward approach (FFN) and the other a nonlinear auto-regressive network (NARX), to predict ET and compared them to the commonly used empirical model Granger and Gray (GG). We trained our models on a nine-year eddy covariance (EC) dataset for Miscanthu s × giganteus ( M . × giganteus ) from Illinois (UIEF), then tested them using out-of-sample data from both UIEF and a different location in Iowa (SABR) to compare the accuracy of FFN, NARX, and GG models in estimating daily ET. A combination of air temperature (T a ) and solar radiation (R s ) was chosen as inputs due to the highest R 2 for FFN (R 2 = 0.79, 0.81, and 0.79 for training, testing, and validation, respectively) and only T a for NARX (R 2 = 0.70 for out-of-sample validation). The predictive power of the FFN model was superior to the NARX and GG models at the UIEF site (R 2 = 0.84, 0.70, and 0.83 for out-of-sample validation, respectively). Our analysis showed that ANN approaches are as accurate as empirical approaches for estimating ET but use fewer inputs.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2026

Dense autoencoders, clustering techniques, and semi-supervised learning for HPGe $γ$-spectra

Classifying high-resolution gamma spectra by their isotopic content is an essential task in nuclear forensics and other applications. Traditional analysis methods are often time-intensive, but machine learning (ML) may help analysts quickly process many spectra. Such methods tend to rely on abundant, well-labeled data for training. Historical gamma data exists in various fields but is not uniformly useful for supervised ML due to inconsistent labeling. Here, to address some of these challenges, we present a method to classify and organize unlabeled data from high-purity germanium detectors using an autoencoding neural network (autoencoder). We trained dense autoencoders to compress gamma data into latent representations that enable efficient data characterization. By clustering the encoded spectra or lower-dimensional mappings of them, we identified and removed portions of over-abundant data categories, resulting in a more balanced dataset and improved autoencoder performance. This encoding and clustering pipeline also enabled the organization of spectra into self-consistent categories. Finally, we found that encoded representations showed potential as inputs for semi-supervised learning of nuclide identification (NID) labels, achieving an average F1 score of 0.85 ± 0.03 when mapping encodings to a set of 65 isotope labels.

Autoencoders↗
DOE OSTI2024

Evolving Competitive Markets in SAPP: Leveraging Competitive Wholesale Electricity Markets to Drive Renewable Generation Capacity in the Southern African Power Pool (SAPP)

The SADC region has significant natural resource potential to increase renewable energy generation, improve electricity reliability, and support economic development. This research finds an apparent lack of confidence from electricity infrastructure investors in SAPP wholesale electricity markets, which increases risk perception and lowers the likelihood of capital deployment. With respect to free market fundamentals, competitive market obstacles and renewable energy development obstacles are characterized. Stakeholders identified the top obstacles to well-functioning competitive markets as insufficient transmission infrastructure for interconnection and regional movement of electricity, dominance of national single-buyer markets, and lack of or weak nation-state regulatory frameworks. Stakeholders prioritized the top three obstacles for renewable energy development as a lack of viable commercial arrangements for variable renewable energy (VRE) balancing, lack of functional and consistent nation-level regulations, and higher project costs related to reliance on imported equipment. With respect to potential solution options, stakeholders prioritized the development of new cost allocation and finance methods to facilitate new transmission expansion, training to educate new or potential new market entrants on SAPP processes, as well as modeling and analysis of regional SAPP participation benefits disaggregated to the nation-state level. From these perspectives, this research identified strategy options for consideration including transitioning SAPP to a regional transmission operator (RTO) for operation and planning of cross-border transmission facilities and market administration, shifting operations of SAPP member transmission systems to Independent System Operators (ISOs), establishing a regional regulatory authority and enhancing market data transparency. Implementing these reforms is expected to be challenging, but not insurmountable, given the domestic political, legal, and jurisdictional complexities of the SADC region.

29 ENERGY PLANNING, POLICY, AND ECONOMY↗
DOE OSTI2023

Finding simplicity: unsupervised discovery of features, patterns, and order parameters via shift-invariant variational autoencoders *

Abstract Recent advances in scanning tunneling and transmission electron microscopies (STM and STEM) have allowed routine generation of large volumes of imaging data containing information on the structure and functionality of materials. The experimental data sets contain signatures of long-range phenomena such as physical order parameter fields, polarization, and strain gradients in STEM, or standing electronic waves and carrier-mediated exchange interactions in STM, all superimposed onto scanning system distortions and gradual changes of contrast due to drift and/or mis-tilt effects. Correspondingly, while the human eye can readily identify certain patterns in the images such as lattice periodicities, repeating structural elements, or microstructures, their automatic extraction and classification are highly non-trivial and universal pathways to accomplish such analyses are absent. We pose that the most distinctive elements of the patterns observed in STM and (S)TEM images are similarity and (almost-) periodicity, behaviors stemming directly from the parsimony of elementary atomic structures, superimposed on the gradual changes reflective of order parameter distributions. However, the discovery of these elements via global Fourier methods is non-trivial due to variability and lack of ideal discrete translation symmetry. To address this problem, we explore the shift-invariant variational autoencoders (shift-VAEs) that allow disentangling characteristic repeating features in the images, their variations, and shifts that inevitably occur when randomly sampling the image space. Shift-VAEs balance the uncertainty in the position of the object of interest with the uncertainty in shape reconstruction. This approach is illustrated for model 1D data, and further extended to synthetic and experimental STM and STEM 2D data. We further introduce an approach for training shift-VAEs that allows finding the latent variables that comport to known physical behavior. In this specific case, the condition is that the latent variable maps should be smooth on the length scale of the atomic lattice (as expected for physical order parameters), but other conditions can be imposed. The opportunities and limitations of the shift VAE analysis for pattern discovery are elucidated.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2023

Synoptic Drivers of Atmospheric River Induced Precipitation Near Dronning Maud Land, Antarctica

Atmospheric rivers (ARs) that reach the Antarctic Ice Sheet (AIS) transport anomalous moisture from lower latitudes and can impact the AIS via extreme precipitation and increased downward longwave radiation. ARs contribute significantly to the interannual variability of precipitation over the AIS and thus are likely to play a key role in understanding future changes in the surface mass balance of the AIS. Dronning Maud Land (DML) is one of four maxima in AR frequency over coastal East Antarctica, with AR precipitation explaining 77% of the interannual variability in precipitation for this region. We employ a 16-node self-organizing map (SOM) trained with MERRA-2 sea-level pressure anomalies to identify synoptic-scale environments associated with landfalling ARs in and around DML. Node composites of atmospheric variables reveal common drivers of precipitation associated with ARs reaching DML including anomalous high-low surface pressure couplets, anomalously high integrated water vapor, and coastal barrier jets. Using a quasi-geostrophic framework, we find that upward vertical motion associated with the occlusion process of attendant cyclones dominates atmospheric lift in AR environments. We further identify mechanisms that explain the variability in AR precipitation intensity across nodes, such as the lift associated with the occlusion process of attendant cyclones and the spatial coincidence of ascent induced by the occlusion process and frontogenesis. Here, the latter suggests that ARs making landfall during the mature phase of cyclogenesis result in higher precipitation intensity compared to landfalling ARs that occur during the occluded phase.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2024

Accuracy, transferability, and computational efficiency of interatomic potentials for simulations of carbon under extreme conditions

Large-scale atomistic molecular dynamics (MD) simulations provide an exceptional opportunity to advance the fundamental understanding of carbon under extreme conditions of high pressures and temperatures. However, the fidelity of these simulations depends heavily on the accuracy of classical interatomic potentials governing the dynamics of many-atom systems. Here, this study critically assesses several popular empirical potentials for carbon, as well as machine learning interatomic potentials (MLIPs), in their ability to simulate a range of physical properties at high pressures and temperatures, including the diamond equation of state, its melting line, shock Hugoniot, uniaxial compressions, and the structure of liquid carbon. Empirical potentials fail to accurately predict the behavior of carbon under high pressure–temperature conditions. In contrast, MLIPs demonstrate quantum accuracy, with Spectral Neighbor Analysis Potential (SNAP) and atomic cluster expansion (ACE) being the most accurate in reproducing the density functional theory results. ACE displays remarkable transferability despite not being specifically trained for extreme conditions. Furthermore, ACE and SNAP exhibit superior computational performance on graphics processing unit-based systems in billion atom MD simulations, with SNAP emerging as the fastest. In addition to offering practical guidance in selecting an interatomic potential with a fine balance of accuracy, transferability, and computational efficiency, this work also highlights transformative opportunities for groundbreaking scientific discoveries facilitated by quantum-accurate MD simulations with MLIPs on emerging exascale supercomputers.

36 MATERIALS SCIENCE↗
DOE OSTI2023

Exhaled breath condensate profiles of U.S. Navy divers following prolonged hyperbaric oxygen (HBO) and nitrogen-oxygen (Nitrox) chamber exposures

Prolonged exposure to hyperbaric hyperoxia can lead to pulmonary oxygen toxicity (PO 2 tox). PO 2 tox is a mission limiting factor for special operations forces divers using closed-circuit rebreathing apparatus and a potential side effect for patients undergoing hyperbaric oxygen (HBO) treatment. In this study, we aim to determine if there is a specific breath profile of compounds in exhaled breath condensate (EBC) that is indicative of the early stages of pulmonary hyperoxic stress/PO 2 tox. Using a double-blind, randomized 'sham' controlled, cross-over design 14 U.S. Navy trained diver volunteers breathed two different gas mixtures at an ambient pressure of 2 ATA (33 fsw, 10 msw) for 6.5 h. One test gas consisted of 100% O 2 (HBO) and the other was a gas mixture containing 30.6% O 2 with the balance N 2 (Nitrox). The high O 2 stress dive (HBO) and low O 2 stress dive (Nitrox) were separated by at least seven days and were conducted dry and at rest inside a hyperbaric chamber. EBC samples were taken immediately before and after each dive and subsequently underwent a targeted and untargeted metabolomics analysis using liquid chromatography coupled to mass spectrometry (LC-MS). Following the HBO dive, 10 out of 14 subjects reported symptoms of the early stages of PO 2 tox and one subject terminated the dive early due to severe symptoms of PO 2 tox. No symptoms of PO 2 tox were reported following the nitrox dive. A partial least-squares discriminant analysis of the normalized (relative to pre-dive) untargeted data gave good classification abilities between the HBO and nitrox EBC with an AUC of 0.99 (±2%) and sensitivity and specificity of 0.93 (±10%) and 0.94 (±10%), respectively. Furthermore, the resulting classifications identified specific biomarkers that included human metabolites and lipids and their derivatives from different metabolic pathways that may explain metabolomic changes resulting from prolonged HBO exposure.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2024

Reinforcement Learning as a Parsimonious Alternative to Prediction Cascades: A Case Study on Image Segmentation

Deep learning architectures have achieved state-of-the-art (SOTA) performance on computer vision tasks such as object detection and image segmentation. This may be attributed to the use of over-parameterized, monolithic deep learning architectures executed on large datasets. Although such large architectures lead to increased accuracy, this is usually accompanied by a larger increase in computation and memory requirements during inference. While this is a non-issue in traditional machine learning (ML) pipelines, the recent confluence of machine learning and fields like the Internet of Things (IoT) has rendered such large architectures infeasible for execution in low-resource settings. For some datasets, large monolithic pipelines may be overkill for simpler inputs. To address this problem, previous efforts have proposed decision cascades where inputs are passed through models of increasing complexity until the desired performance is achieved. However, we argue that cascaded prediction leads to sub-optimal throughput and increased computational cost due to wasteful intermediate computations. To address this, we propose PaSeR (Parsimonious Segmentation with Reinforcement Learning) a non-cascading, cost-aware learning pipeline as an efficient alternative to cascaded decision architectures. Through experimental evaluation on both real-world and standard datasets, we demonstrate that PaSeR achieves better accuracy while minimizing computational cost relative to cascaded models. Further, we introduce a new metric IoU/GigaFlop to evaluate the balance between cost and performance. On the real-world task of battery material phase segmentation, PaSeR yields 179% improvement over SOTA MatPhase model and a 196% improvement over IDK Cascades under the IoU/GigaFlop metric. We also demonstrate PaSeR’s adaptability to complementary models trained on a noisy MNIST dataset, where it outperforms all baselines on IoU/GigaFlop by an average of 44%.

Srikshan, Bharat↗
DOE OSTI2024

Effects of input gradient regularization on neural networks time-series forecasting of thermal power systems

This study proposes using neural networks, specifically gated recurrent unit (GRU), long-short-term memory (LSTM), and transformer networks, to improve control strategies in a 450 MW coal-fired power plant. However, neural networks face issues of becoming overly dependent on just a few variables to make predictions, which negatively impacts control decisions that rely on the model to determine the value of all manipulated variables. The paper introduces regularization techniques, including noise injection and input gradient regularization, during the training phase. Here, the work presents novel contributions in adapting neural networks to control industrial systems and applying regularization techniques from computer vision to industrial process control. Results demonstrate the effectiveness of input gradient regularization in reducing model dependence on subsets of variables, emphasizing the balance between fidelity and controllability. Further exploration is recommended, including the development of recurrent transformers, closed-loop control testing, and a sensitivity analysis on computer models to provide further insight.

20 FOSSIL-FUELED POWER PLANTS↗