Search NASA⌕ Search

SEARCH · Search NASA

Results for “adaptive learning”

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 595 records · Page 33

Unsupervised domain adaptation for radioisotope identification in gamma spectroscopy

Training machine learning models for radioisotope identification using gamma spectroscopy remains an elusive challenge for many practical applications, largely stemming from the difficulty of acquiring and labeling large, diverse experimental datasets. Simulations can mitigate this challenge, but the accuracy of models trained on simulated data can deteriorate substantially when deployed to an out-of-distribution operational environment. In this study, we demonstrate that unsupervised domain adaptation (UDA) can improve the ability of a model trained on synthetic data to generalize to a new testing domain, provided unlabeled data from the target domain are available. Conventional supervised techniques are unable to utilize this data because the absence of isotope labels precludes defining a supervised classification loss. Instead, we first pretrain a spectral classifier using labeled synthetic data and subsequently leverage unlabeled target data to align the learned feature representations between the source and target domains. We compare a range of different UDA techniques, finding that minimizing the maximum mean discrepancy (MMD) between source and target feature vectors yields the most consistent improvement to testing scores. For instance, using a custom transformer-based neural network, we achieved a testing accuracy of $0.904 \pm 0.022$ on an experimental LaBr test set after performing unsupervised feature alignment via MMD minimization, compared to $0.754 \pm 0.014$ before alignment. Overall, our results highlight the potential of using UDA to adapt a radioisotope classifier trained on synthetic data for real-world deployment.

Lalor, Peter W.↗

Automatic Detection and Classification of Aurora in THEMIS All‐Sky Images

We report a novel machine-learning algorithm for automatically detecting and classifying aurora in all–sky images (ASI) that is largely trained without requiring ground–truth labels. By including a small number of labeled images, we are able to automatically label all of the approximately 700 million images in the Time History of Events and Macroscale Interactions during Substorms (THEMIS) ASI data set from 2008 to 2022. We use a two–stage approach. In the first stage, we adapt the Simple framework for Contrastive Learning of Representations (SimCLR) algorithm to learn latent representations of THEMIS all–sky images. We then finetune a classifier network on the latent representations our model learns of the manually labeled Oslo aurora THEMIS (OATH) data set. We demonstrate that this two–stage approach achieves excellent classification results on data for which there is no current ML classification benchmark. The outcome of this work will facilitate efficient information retrieval for researchers interested in specific categories of aurora and will enable large scale statistical studies and machine learning analyses of THEMIS all–sky images that have not previously been possible. To demonstrate possible ways to utilize this database, we performed a statistical analysis of the occurrence rates of auroral labels with respect to solar wind parameters, interplanetary magnetic field vector, and geomagnetic indices. We further investigate the occurrence rates of auroral phenomena in the annotated data set and their geoeffectiveness by utilizing the co–located THEMIS ground magnetometer data set.

Jeremiah W Johnson↗

Mimicking associative learning using an ion-trapping non-volatile synaptic organic electrochemical transistor

Associative learning, a critical learning principle to improve an individual’s adaptability, has been emulated by few organic electrochemical devices. However, complicated bias schemes, high write voltages, as well as process irreversibility hinder the further development of associative learning circuits. Here, by adopting a poly(3,4-ethylenedioxythiophene):tosylate/Polytetrahydrofuran composite as the active channel, we present a non-volatile organic electrochemical transistor that shows a write bias less than 0.8 V and retention time longer than 200 min without decoupling the write and read operations. By incorporating a pressure sensor and a photoresistor, a neuromorphic circuit is demonstrated with the ability to associate two physical inputs (light and pressure) instead of normally demonstrated electrical inputs in other associative learning circuits. To unravel the non-volatility of this material, ultraviolet-visible-near-infrared spectroscopy, X-ray photoelectron spectroscopy and grazing-incidence wide-angle X-ray scattering are used to characterize the oxidation level variation, compositional change, and the structural modulation of the poly(3,4-ethylenedioxythiophene):tosylate/Polytetrahydrofuran films in various conductance states. The implementation of the associative learning circuit as well as the understanding of the non-volatile material represent critical advances for organic electrochemical devices in neuromorphic applications.

42 ENGINEERING↗

A Learning-based Power Management Method for Networked Microgrids Under Incomplete Information

In this paper, we present a model-free reinforcement learning (RL)-based technique to solve the power management problem of networked microgrids (MGs), when the utility has only limited or no detailed knowledge of the MG asset models.The proposed method consists of two levels: at the first level, using an approximate RL strategy and aggregated data, a utility agent determines the locational price signals to maximize its profit under incomplete information on the MGs’ private data behind their points of common coupling (PCC). The proposed RL strategy enables the utility to adapt to changing system conditions and learn from previous experiences, while respecting the data ownership of the MGs. At the second level, the MGs receive the price signals to dispatch their generation/storage assets individually over look-ahead decision time window, while considering power flow constraints. The performance of the proposed RL technique has been verified under realistic operational scenarios.

Zhang, Qianzhi↗

Parallelization of a Six Degree of Freedom Entry Vehicle Trajectory Simulation Using OpenMP and OpenACC

The art and science of writing parallelized software, using methods such as Open Multi-Processing (OpenMP) and Open Accelerators (OpenACC), is dominated by computer scientists. Engineers and non-computer scientists looking to apply these techniques to their project applications face a steep learning curve, especially when looking to adapt their original single threaded software to run multi-threaded on graphics processing units (GPUs). There are significant changes in mindset that must occur; such as how to manage memory, the organization of instructions, and the use of if statements (also known as branching). The purpose of this work is twofold: 1) to demonstrate the applicability of parallelized coding methodologies, OpenMP and OpenACC, to tasks outside of the typical large scale matrix mathematics; and 2) to discuss, from an engineer’s perspective, the lessons learned from parallelizing software using these computer science techniques. This work applies OpenMP, on both multi-core central processing units (CPUs) and Intel® Xeon Phi™ 7210, and OpenACC on GPUs. These parallelization techniques are used to tackle the simulation of thousands of entry vehicle trajectories through the integration of six degree of freedom (DoF) equations of motion (EoM). The forces and moments acting on the entry vehicle, and used by the EoM, are estimated using multiple models of varying levels of complexity. Several benchmark comparisons are made on the execution of six DoF trajectory simulation: single thread Intel® Xeon® E5-2670 CPU, multi-thread CPU using OpenMP, multi-thread Xeon Phi™ 7210 using OpenMP, and multi-thread NVIDIA® Tesla® K40 GPU using OpenACC. These benchmarks are run on the Pleiades Supercomputer Cluster at the National Aeronautics and Space Administration (NASA) Ames Research Center (ARC), and a Xeon Phi™ 7210 node at NASA Langley Research Center (LaRC).

Green, Justin S.↗

A neural fuzzy controller learning by fuzzy error propagation

In this paper, we describe a procedure to integrate techniques for the adaptation of membership functions in a linguistic variable based fuzzy control environment by using neural network learning principles. This is an extension to our work. We solve this problem by defining a fuzzy error that is propagated back through the architecture of our fuzzy controller. According to this fuzzy error and the strength of its antecedent each fuzzy rule determines its amount of error. Depending on the current state of the controlled system and the control action derived from the conclusion, each rule tunes the membership functions of its antecedent and its conclusion. By this we get an unsupervised learning technique that enables a fuzzy controller to adapt to a control task by knowing just about the global state and the fuzzy error.

Nauck, Detlef↗

An adaptive sampling augmented Lagrangian method for stochastic optimization with deterministic constraints

The primary goal of this paper is to provide an efficient solution algorithm based on the augmented Lagrangian framework for optimization problems with a stochastic objective function and deterministic constraints. Our main contribution is combining the augmented Lagrangian framework with adaptive sampling, resulting in an efficient optimization methodology validated with practical examples. To achieve the presented efficiency, here we consider inexact solutions for the augmented Lagrangian subproblems, and through an adaptive sampling mechanism, we control the variance in the gradient estimates. Furthermore, we analyze the theoretical performance of the proposed scheme by showing equivalence to a gradient descent algorithm on a Moreau envelope function, and we prove sublinear convergence for convex objectives and linear convergence for strongly convex objectives with affine equality constraints. The worst-case sample complexity of the resulting algorithm, for an arbitrary choice of penalty parameter in the augmented Lagrangian function, is $\mathscr{O}$(ϵ -3-δ ) , where ϵ > 0 is the expected error of the solution and δ > 0 is a user-defined parameter. If the penalty parameter is chosen to be $\mathscr{O}$(ϵ -1 ), we demonstrate that the result can be improved to $\mathscr{O}$(ϵ -2 ) , which is competitive with the other methods employed in the literature. Moreover, if the objective function is strongly convex with affine equality constraints, we obtain $\mathscr{O}$(ϵ -1 log(1/ϵ)) complexity. Finally, we empirically verify the performance of our adaptive sampling augmented Lagrangian framework in machine learning optimization and engineering design problems, including topology optimization of a heat sink with environmental uncertainty.

97 MATHEMATICS AND COMPUTING↗

Spaceflight Sensorimotor Analogs: Simulating Acute and Adaptive Effects

Adaptive changes in sensorimotor function during spaceflight are reflected by spatial disorientation, motion sickness, gaze destabilization and decrements in balance, locomotion and eye-hand coordination that occur during and following transitions between different gravitational states. The purpose of this study was to conduct a meta-synthesis of data from spaceflight analogs to evaluate their effectiveness in simulating adaptive changes in sensorimotor function. METHODS. The analogs under review were categorized as either acute analogs used to simulate performance decrements accompanied with transient changes, or adaptive analogs used to drive sensorimotor learning to altered sensory feedback. The effectiveness of each analog was evaluated in terms of mechanisms of action, magnitude and time course of observed deficits compared to spaceflight data, and the effects of amplitude and exposure duration. RESULTS. Parabolic flight has been used extensively to examine effects of acute variation in gravitational loads, ranging from hypergravity to microgravity. More recently, galvanic vestibular stimulation has been used to elicit acute postural, locomotor and gaze dysfunction by disrupting vestibular afferents. Patient populations, e.g., with bilateral vestibular loss or cerebellar dysfunction, have been proposed to model acute sensorimotor dysfunction. Early research sponsored by NASA involved living onboard rotating rooms, which appeared to approximate the time course of adaptation and post-exposure recovery observed in astronauts following spaceflight. Exposure to different bed-rest paradigms (6 deg head down, dry immersion) result in similar motor deficits to that observed following spaceflight. Shorter adaptive analogs have incorporated virtual reality environments, visual distortion paradigms, exposure to conflicting tilt-translation cues, and exposure to 3Gx centrifugation. As with spaceflight, there is considerable variability in responses to most of the analogs reviewed. DISCUSSION. A true ground-based flight analog for sensorimotor function is not feasible. A combination of flight analogs; however, can be used to selectively mimic different aspects of the spaceflight-induced sensorimotor performance decrements.

Taylor, Laura C.↗

Ensuring Success of Adaptive Control Research Through Project Lifecycle Risk Mitigation

Lessons Learne: 1. Design-out unnecessary risk to prevent excessive mitigation management during flight. 2. Consider iterative checkouts to confirm or improve human factor characteristics. 3. Consider the total flight test profile to uncover unanticipated human-algorithm interactions. 4. Consider test card cadence as a metric to assess test readiness. 5. Full-scale flight test is critical to development, maturation, and acceptance of adaptive control laws for operational use.

Pavlock, Kate M.↗

The Effect of Spaceflight on the Ultrastructure of the Cerebellum

In weightlessness, astronauts and cosmonauts may experience postural illusions as well as motion sickness symptoms known as the space adaptation syndrome. Upon return to Earth, they have irregularities in posture and balance. The adaptation to microgravity and subsequent re-adaptation to Earth occurs over several days. At the cellular level, a process called neuronal plasticity may mediate this adaptation. The term plasticity refers to the flexibility and modifiability in the architecture and functions of the nervous system. In fact, plastic changes are thought to underlie not just behavioral adaptation, but also the more generalized phenomena of learning and memory. The goal of this experiment was to identify some of the structural alterations that occur in the rat brain during the sensory and motor adaptation to microgravity. One brain region where plasticity has been studied extensively is the cerebellar cortex-a structure thought to be critical for motor control, coordination, the timing of movements, and, most relevant to the present experiment, motor learning. Also, there are direct as well as indirect connections between projections from the gravity-sensing otolith organs and several subregions of the cerebellum. We tested the hypothesis that alterations in the ultrastructural (the structure within the cell) architecture of rat cerebellar cortex occur during the early period of adaptation to microgravity, as the cerebellum adapts to the absence of the usual gravitational inputs. The results show ultrastructural evidence for neuronal plasticity in the central nervous system of adult rats after 24 hours of spaceflight. Qualitative studies conducted on tissue from the cerebellar cortex (specifically, the nodulus of the cerebellum) indicate that ultrastructural signs of plasticity are present in the cerebellar zones that receive input from the gravity-sensing organs in the inner ear (the otoliths). These changes are not observed in this region in cagematched ground control animals. The specific changes include the formation of lamellar bodies, profoundly enlarged Purkinje cell mitochondria, the presence of inter-neuronal cellular protrusions in the molecular layer, and signs of degeneration in the distal dendrites of the Purkinje cells. Since these morphologic signs are not apparent in the control animals, they are not likely to be due to caging or tissue processing effects. The particular nature of the structural alterations in the nodulus, most notably the formation of lamellar bodies and the presence of degeneration, further suggests that excitotoxicity (damaging overstimulation of neurons) may play a role in the short-term neural response to spaceflight. These findings suggest a structural basis for the neuronal and synaptic plasticity accompanying the central nervous system response to altered gravity and help identify the cellular bases underlying the vestibular abnormalities experienced by astronauts during periods of adaptation and re-adaptation to different gravitational forces. Also, since the short- and long-term changes in neural structure occurring during such periods of adaptation resemble the neuronal alterations that occur in some neurologic disorders such as stroke, these findings may offer guidance in the development of strategies for rehabilitation and treatment of such disorders.

Holstein, Gay R.↗

Change detection and classification of land cover in multispectral satellite imagery using clustering of sparse approximations (CoSA) over learned feature dictionaries

Neuromimetic machine vision and pattern recognition algorithms are of great interest for landscape characterization and change detection in satellite imagery in support of global climate change science and modeling. We present results from an ongoing effort to extend machine vision methods to the environmental sciences, using adaptive sparse signal processing combined with machine learning. A Hebbian learning rule is used to build multispectral, multiresolution dictionaries from regional satellite normalized band difference index data. Land cover labels are automatically generated via our CoSA algorithm: Clustering of Sparse Approximations, using a clustering distance metric that combines spectral and spatial textural characteristics to help separate geologic, vegetative, and hydrologie features. We demonstrate our method on example Worldview-2 satellite images of an Arctic region, and use CoSA labels to detect seasonal surface changes. In conclusion, our results suggest that neuroscience-based models are a promising approach to practical pattern recognition and change detection problems in remote sensing.

97 MATHEMATICS AND COMPUTING↗

SIDDA: SInkhorn Dynamic Domain Adaptation

Modern neural networks (NNs) often do not generalize well in the presence of a "covariate shift"; that is, in situations where the training and test data distributions differ, but the conditional distribution of classification labels remains unchanged. In such cases, NN generalization can be reduced to a problem of learning more domain-invariant features. Domain adaptation (DA) methods include a range of techniques aimed at achieving this; however, these methods have struggled with the need for extensive hyperparameter tuning, which then incurs significant computational costs. In this work, we introduce SIDDA, an out-of-the-box DA training algorithm built upon the Sinkhorn divergence, that can achieve effective domain alignment with minimal hyperparameter tuning and computational overhead. We demonstrate the efficacy of our method on multiple simulated and real datasets of varying complexity, including simple shapes, handwritten digits, and real astronomical observations. SIDDA is compatible with a variety of NN architectures, and it works particularly well in improving classification accuracy and model calibration when paired with equivariant neural networks (ENNs). We find that SIDDA enhances the generalization capabilities of NNs, achieving up to a ≈40% improvement in classification accuracy on unlabeled target data. We also study the efficacy of DA on ENNs with respect to the varying group orders of the dihedral group DN, and find that the model performance improves as the degree of equivariance increases. Finally, we find that SIDDA enhances model calibration on both source and target data--achieving over an order of magnitude improvement in the ECE and Brier score. SIDDA's versatility, combined with its automated approach to domain alignment, has the potential to advance multi-dataset studies by enabling the development of highly generalizable models.

Pandya, Sneh [Northeastern U.]↗

Which Way is Up? Lessons Learned from Space Shuttle Sensorimotor Research

The Space Shuttle Program provided the opportunity to examine sensorimotor adaptation to space flight in unprecedented numbers of astronauts, including many over multiple missions. Space motion sickness (SMS) severity was highly variable across crewmembers. SMS generally lasted 2-3 days in-flight with approximately 1/3 of crewmembers experiencing moderate to severe symptoms, and decreased incidence in repeat flyers. While SMS has proven difficult to predict from susceptibility to terrestrial analogs, symptoms were alleviated by medications, restriction of early activities, maintaining familiar orientation with respect to the visual environment and maintaining contact cues. Adaptive changes were also reflected by the oculomotor and perceptual disturbances experienced early inflight and by the perceptual and motor coordination problems experienced during re-entry and landing. According to crew self-reports, systematic head movements performed during reentry, as long as paced within one's threshold for motion tolerance, facilitated the early readaptation process. The Shuttle provided early postflight crew access to document the initial performance decrements and time course of recovery. These early postflight measurements were critical to inform the program of risks associated with extending the duration of Shuttle missions. Neurological postflight deficits were documented using a standardized subjective rating by flight surgeons. Computerized dynamic posturography was also implemented as a quantitative means of assessing sensorimotor function to support crew return-to-duty assessments. Towards the end of the Shuttle Program, more emphasis has been placed on mapping physiological changes to functional performance. Future commercial flights will benefit from pre-mission training including exposures to launch and entry G transitions and sensorimotor adaptability assessments. While SMS medication usage will continue to be refined, non-pharmacological countermeasures (e.g., sensory aids) will have both space and Earth-based applications. Early postflight field tests are recommended to provide the evidence base for best practices for future commercial flight programs. Learning Objective: Overview of the Space Shuttle Program regarding adaptive changes in sensorimotor function, including what was learned from research, what was implemented for medical operations, and what is recommended for commercial flights.

Wood, S. J.↗

Intelligent flight control systems

The capabilities of flight control systems can be enhanced by designing them to emulate functions of natural intelligence. Intelligent control functions fall in three categories. Declarative actions involve decision-making, providing models for system monitoring, goal planning, and system/scenario identification. Procedural actions concern skilled behavior and have parallels in guidance, navigation, and adaptation. Reflexive actions are spontaneous, inner-loop responses for control and estimation. Intelligent flight control systems learn knowledge of the aircraft and its mission and adapt to changes in the flight environment. Cognitive models form an efficient basis for integrating 'outer-loop/inner-loop' control functions and for developing robust parallel-processing algorithms.

Stengel, Robert F.↗

Simulation Based Inference with Domain Adaptation for Strong Gravitational Lensing

Simulation based inference leverages machine learning to carry out Bayesian inference for systems with intractable likelihoods. However, transitioning a network trained on simulated data to real data runs the risk of encountering domain shift, leading to performance losses. We attempt to implement domain adaptation into the sbi neural posterior estimation framework using the Maximum Mean Discrepancy as an additional network loss. We test two network architectures and use masked autoregressive flow for density estimation. We test the network on a set of 400,000 simulated strong gravitational lensing images generated using deeplenstronomy. The source domain is defined as low noise whereas the target domain has a noise profile sampled from experimentally derived DES survey conditions. We find that, while DA does lead to performance improvements, they are marginal at ~6% less inference error. We also find a similar marginal improvement in uncertainty calibration at around 8%.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

Simulation based inference with domain adaptation for strong gravitational lensing

Simulation based inference leverages machine learning to carry out Bayesian inference in systems with intractable likelihoods. However, transitioning a network trained on simulated data to real data runs the risk of encountering domain shift, leading to performance losses. We attempt to implement domain adaptation into the sbi neural posterior estimation framework using the Maximum Mean Discrepancy as an additional network loss, using masked autoregressive fow (MAF) as our density estimator. We test the network on a set of 400,000 simulated strong gravitational lensing images generated using deeplenstronomy. The source domain is defined as low noise whereas the target domain has a noise profile sampled from experimentally derived DES survey conditions. We find that SBI appears robust against small changes in the data with similar performance on source and target. Moreover, while DA does lead to performance improvements, they are marginal at 6% less inference error.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

PMDT: AI-Enabled Predictive Maintenance Digital Twins for Advanced Nuclear Reactors

Our team made substantial technical progress on various fronts during the course of the program. Multiple milestones were geared towards demonstrating the feasibility of machine learning based predictive maintenance digital twins towards reducing O&M costs, whereas some other milestones actually focused on identifying technical gaps and developing technologies such as humble AI to provide necessary robustness to the ML-based models. We were able to demonstrate in many cases that Machine learning-based methods can be successfully adapted for Nuclear plant environments especially for remote monitoring applications. Detailed analyses were carried out with plant and full scope simulation data along with capabilities of enhanced analytics to assess and set realistic expectations on cost reductions in O&M. These assessments are paving the way for investments towards reactor design improvements as well project planning for SMR projects as they develop and mature in the next few years. Technology developed under this program got direct visibility to GE Hitachi and their utility customers and resulted in positive intents to deploy some of the elements from design phase. The project additionally resulted in several reports, publications, software and data generation that will be useful in deployment and O&M services for BWRX300 fleets.

21 SPECIFIC NUCLEAR REACTORS AND ASSOCIATED PLANTS↗

The Applied Meteorology Unit: Nineteen Years Successfully Transitioning Research into Operations for America's Space Program

The Applied Meteorology Unit (AMU) provides technology transition and technique development to improve operational weather support to the Space Shuttle and the entire American space program. The AMU is funded and managed by NASA and operated by a contractor that provides five meteorologists with a diverse mix of advanced degrees, operational experience, and associated skills including data processing, statistics, and the development of graphical user interfaces. The AMU's primary customers are the U.S. Air Force 45th Weather Squadron at Patrick Air Force Base, the National Weather Service Spaceflight Meteorology Group at NASA Johnson Space Center, and the National Weather Service Melbourne FL Forecast Office. The AMU has transitioned research into operations for nineteen years and worked on a wide range of topics, including new forecasting techniques for lightning probability, synoptic peak winds,.convective winds, and summer severe weather; satellite tools to predict anvil cloud trajectories and evaluate camera line of sight for Space Shuttle launch; optimized radar scan strategies; evaluated and implemented local numerical models; evaluated weather sensors; and many more. The AMU has completed 113 projects with 5 more scheduled to be completed by the end of 2010. During this rich history, the AMU and its customers have learned many lessons on how to effectively transition research into operations. Some of these lessons learned include collocating with the operational customer and periodically visiting geographically separated customers, operator submitted projects, consensus tasking process, use of operator primary advocates for each project, customer AMU liaisons with experience in both operations and research, flexibility in adapting the project plan based on lessons learned during the project, and incorporating training and other transition assistance into the project plans. Operator involvement has been critical to the AMU's remarkable success and many awards from NASA, the National Weather Association, and two citations from the Navy's Center of Excellence for Best Manufacturing Practices. This paper will present the AMU's proven methods and explain how they may be applied by other organizations to effectively transition research into operations.

Madura, John T.↗