SEARCH · Search NASA

Results for “Continual 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.

SourceResearch connectionFrom yearThrough year

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 109 records · Page 6

All sources NASA NTRS DOE OSTI

DOE OSTI2024

Calibrating Bayesian generative machine learning for Bayesiamplification

Recently, combinations of generative and Bayesian deep learning have been introduced in particle physics for both fast detector simulation and inference tasks. These neural networks aim to quantify the uncertainty on the generated distribution originating from limited training statistics. The interpretation of a distribution-wide uncertainty however remains ill-defined. We show a clear scheme for quantifying the calibration of Bayesian generative machine learning models. For a Continuous Normalizing Flow applied to a low-dimensional toy example, we evaluate the calibration of Bayesian uncertainties from either a mean-field Gaussian weight posterior, or Monte Carlo sampling network weights, to gauge their behaviour on unsteady distribution edges. Well calibrated uncertainties can then be used to roughly estimate the number of uncorrelated truth samples that are equivalent to the generated sample and clearly indicate data amplification for smooth features of the distribution.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2024

Advancing set-conditional set generation: Diffusion models for fast simulation of reconstructed particles

The computational intensity of detector simulation and event reconstruction poses a significant difficulty for data analysis in collider experiments. This challenge inspires the continued development of machine learning techniques to serve as efficient surrogate models. We propose a fast emulation approach that combines simulation and reconstruction. In other words, a neural network generates a set of reconstructed objects conditioned on input particle sets. To make this possible, we advance set-conditional set generation with diffusion models. Using a realistic, generic, and public detector simulation and reconstruction package (COCOA), we show how diffusion models can accurately model the complex spectrum of reconstructed particles inside jets.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

DOE OSTI2025

Deep Reinforcement Learning Based Control of Wind Turbines for Fast Frequency Response

In order to fulfill vital auxiliary grid services, such as load regulation, spin and non-spin reserve provision, and frequency support during emergencies, there is often a requirement for certain wind farms to operate in de-loaded modes. Leveraging the swift response capabilities of wind farms, this study demonstrates that reserving power in de-loaded modes can significantly enhance power grid stability and reliability during system contingencies. Controlling wind farms optimally for frequency support is intricate due to the nonlinearity of models and controllers and the complexity of wind farm interactions with power systems. Here, to address this challenge, this paper introduces a novel approach that integrates wind turbines into reinforcement learning-based solutions for frequency response. This innovative methodology utilizes the state-of-the-art reinforcement learning algorithm known as the surrogate-gradient-based evolutionary strategy. The proposed learning-based algorithm provides continuous control of wind farm output to rapidly stabilize system frequency and prevent unnecessary trips of under-frequency load shedding relays. To facilitate efficient training, parallel computing techniques are employed. The proposed methodology is evaluated on a modified IEEE-39 bus system, and simulation results reveal its efficacy in reliably supporting power system frequency and preventing the need for unnecessary load shedding.

Gao, Wei [Argonne National Laboratory (ANL), Argon↗

DOE OSTI2025

Mathematical Morphological Filtering with a Self-Adaptive Reconstruction Technique and Application to Local Seismic Data

Recorded seismic data are generally contaminated by noise from different sources, which masks the signals of interest. In the seismology community, frequency filtering (FF) is the standard method for noise suppression. However, when the signal of interest and noise share the same frequency band, the latter cannot be filtered out without infringing on the former. We implemented a noise suppression approach based on the mathematical morphology theorem. The method involves compound operations of dilation and erosion using structuring elements of varying lengths and decomposes an input noisy waveform into several time functions with differing characteristics. Further, the filtered waveform is constructed from the time functions using a self-adaptive reconstruction technique. Application to a data set of >4700 local waveforms suggests that the implemented mathematical morphological filtering (MMF) approach is efficient for data with low signal-to-noise ratio (SNR) and significantly outperforms FF in that SNR range. For most of the dataset, FF, machine learning (ML) denoising, and continuous wavelet transform (CWT) thresholding result in higher SNR values compared with the MMF method. However, for ~42% of the waveforms, MMF outperforms FF, and the SNR gain achieved with MMF is as large as ~23 dB. Compared to ML denoising and CWT thresholding, this proportion drops to only ~10%–14%. Our results suggests that in an operational setting, MMF cannot replace the other noise suppression methods; however, signal detection can be improved if MMF is used to supplement them in some scenarios. MMF could help detect signals in problematic low-SNR data, which are currently being missed particularly when using FF alone.

58 GEOSCIENCES↗

NASA NTRS1986

Learned pattern recognition using synthetic-discriminant-functions

A method of using synthetic-discriminant-functions to facilitate learning in a pattern recognition system is discussed. Learning is accomplished by continually adding images to the training set used for synthetic discriminant functions (SDF) construction. Object identification is performed by efficiently searching a library of SDF filters for the maximum optical correlation. Two library structures are discussed - binary tree and multilinked graph - along with maximum ascent, back-tracking, perturbation, and simulated annealing searching techniques. By incorporating the distortion invariant properties of SDFs within a library structure, a robust pattern recognition system can be produced.

Jared, David A.↗

NASA NTRS1987

Information transfer in pilots' use of a collision avoidance system

A flight simulator study of pilots' use of the Traffic-alert and Collision Avoidance System is described. Three levels of information on the location of other air traffic were presented to different groups of airline pilots. The amount of informtion on the location of other traffic had little effect on pilots' performance of the maneuvers commanded by the collision avoidance system. Measured crew responses were similar with no presentation of traffic location, with limited information, and with continuous traffic information. No learning effects were observed, and differences in flight experience did not contribute to the performance difference found.

Chappell, Sheryl L.↗

NASA NTRS2005

Distributed Adaptive Control: Beyond Single-Instant, Discrete Variables

In extensive form noncooperative game theory, at each instant t, each agent i sets its state x, independently of the other agents, by sampling an associated distribution, q(sub i)(x(sub i)). The coupling between the agents arises in the joint evolution of those distributions. Distributed control problems can be cast the same way. In those problems the system designer sets aspects of the joint evolution of the distributions to try to optimize the goal for the overall system. Now information theory tells us what the separate q(sub i) of the agents are most likely to be if the system were to have a particular expected value of the objective function G(x(sub 1),x(sub 2), ...). So one can view the job of the system designer as speeding an iterative process. Each step of that process starts with a specified value of E(G), and the convergence of the q(sub i) to the most likely set of distributions consistent with that value. After this the target value for E(sub q)(G) is lowered, and then the process repeats. Previous work has elaborated many schemes for implementing this process when the underlying variables x(sub i) all have a finite number of possible values and G does not extend to multiple instants in time. That work also is based on a fixed mapping from agents to control devices, so that the the statistical independence of the agents' moves means independence of the device states. This paper also extends that work to relax all of these restrictions. This extends the applicability of that work to include continuous spaces and Reinforcement Learning. This paper also elaborates how some of that earlier work can be viewed as a first-principles justification of evolution-based search algorithms.

Wolpert, David H.↗

NASA NTRS2008

Building on the Past - Looking to the Future: A Focus on Payload Safety

The history of the space industry stretches far and above lunar landings to the construction of the International Space Station. For years, humans have sought to understand the nature of the universe. As society grows in knowledge and curiosity of space, the focus of maintaining the safety of the crew and vehicle habitability is of utmost importance to the National Aeronautics and Space Administration (NASA) community. Through the years, Payload Safety has developed not only as a Panel, but also as part of the NASA community, striving to enhance the efficiency and understanding of how business should be conducted as more International Partners become involved. This is the first in a series of papers and presentations in what is hoped to be an annual update that provides continuous challenges and lessons learned in the areas of communication, safety requirements and processes and other areas which have been vital to the Payload Safety Review Panel (PSRP).

Nash, Sally K.↗

NASA NTRS2013

Improved CPAS Photogrammetric Capabilities for Engineering Development Unit (EDU) Testing

This paper focuses on two key improvements to the photogrammetric analysis capabilities of the Capsule Parachute Assembly System (CPAS) for the Orion vehicle. The Engineering Development Unit (EDU) system deploys Drogue and Pilot parachutes via mortar, where an important metric is the muzzle velocity. This can be estimated using a high speed camera pointed along the mortar trajectory. The distance to the camera is computed from the apparent size of features of known dimension. This method was validated with a ground test and compares favorably with simulations. The second major photogrammetric product is measuring the geometry of the Main parachute cluster during steady-state descent using onboard cameras. This is challenging as the current test vehicles are suspended by a single-point attachment unlike earlier stable platforms suspended under a confluence fitting. The mathematical modeling of fly-out angles and projected areas has undergone significant revision. As the test program continues, several lessons were learned about optimizing the camera usage, installation, and settings to obtain the highest quality imagery possible.

Ray, Eric S.↗

NASA NTRS2018

Preparing to Assure Mission Success

Space flight missions are inherently risky. The mission operations assurance discipline grew out of strategies to control error and evolved into what today is an independent technical authority dedicated to achieving mission success. After the Mars Climate Orbiter loss, NASA made a set of recommendations for missions. Two of those recommendations led to the creation of today’s Mission Operations Assurance Manager role: (1) require an independent Mission Assurance representative during the operational phase of every flight project, and (2) require all flight projects to report and track post-launch anomalies. Since then, the MOAM role has been continually refined by lessons learned from its practice on more than 20 concurrent missions. The MOAM role requires significant operations experience as well as an additional set of skills. In the interests of developing a cadre of potential MOAMs, the MOAM training program was developed. It is a 14-unit set of presentations/discussions covering the 11 categories of MOAM tasks. It is offered to new MOAMs and interested systems engineers with an eye to preparing both types of personnel for present and future assignments, as well as building relationships among personnel of different organizations. MOAM training sessions include presentations on the various tools and processes employed by MOAMs, interspersed with mindset lessons illustrated via “war story” discussions of previous missions.

Bryant, Larry↗

NASA NTRSDate not supplied

Managing Maintenance Error: Six Lessons From Aviation Maintenance

There are clear parallels between the ground processing of spacecraft and the maintenance of airline aircraft. In both cases, reliable human performance is critical to safe outcomes. Examples of human factors in airline maintenance and spacecraft ground processing illustrate the similarities between these two domains. The worldwide aviation industry began to pay close attention to human factors in maintenance after several maintenance-related disasters in the 1970s and 80s. Rather than simply applying solutions that had been developed for flight crew and air traffic controllers, the aviation industry developed human factors interventions specifically tailored for maintenance personnel. These interventions have led to safety improvements, greater reliability, and significant cost savings. Six human factors approaches from airline maintenance that can be usefully applied to enhance the quality and safety of ground processing are outlined. These are 1) Design for assembly, test, and maintenance, 2) Non technical skills training, 3) Improved design of documentation, 4) Reduction of iatrogenic quality lapses, 5) Barrier and control analysis, and 6) Continuous improvement based on learning from quality lapses.

maintenance human factors↗

DOE OSTI2026

TANTE: Time-adaptive operator learning via neural Taylor expansion

Operator learning for time-dependent partial differential equations (PDEs) has seen rapid progress in recent years, enabling efficient approximation of complex spatiotemporal dynamics. However, most existing methods rely on fixed time step sizes during rollout, which limits their ability to adapt to varying temporal complexity and often leads to error accumulation. In this work, we propose the Time-Adaptive Transformer with Neural Taylor Expansion (TANTE), a novel operator-learning framework that produces continuous-time predictions with adaptive step sizes. TANTE predicts future states by performing a Taylor expansion at the current state, where neural networks learn both the higher-order temporal derivatives and the local radius of convergence. This allows the model to dynamically adjust its rollout based on the local behavior of the solution, thereby reducing cumulative error and improving computational efficiency. We demonstrate the effectiveness of TANTE across a wide range of PDE benchmarks, achieving superior accuracy and adaptability compared to fixed-step baselines, delivering accuracy gains of 60-80 % and speed-ups of 30-40 % at inference time.

97 MATHEMATICS AND COMPUTING↗

NASA NTRS1995

Building an adaptive agent to monitor and repair the electrical power system of an orbital satellite

Over several years we have developed a multistrategy apprenticeship learning methodology for building knowledge-based systems. Recently we have developed and applied our methodology to building intelligent agents. This methodology allows a subject matter expert to build an agent in the same way in which the expert would teach a human apprentice. The expert will give the agent specific examples of problems and solutions, explanations of these solutions, or supervise the agent as it solves new problems. During such interactions, the agent learns general rules and concepts, continuously extending and improving its knowledge base. In this paper we present initial results on applying this methodology to build an intelligent adaptive agent for monitoring and repair of the electrical power system of an orbital satellite, stressing the interaction with the expert during apprenticeship learning.

Tecuci, Gheorghe↗

NASA NTRS2020

Amino Acid Encoding for Deep Learning Applications

Background: The number of applications of deep learning algorithms in bioinformatics is increasing as they usually achieve superior performance over classical approaches, especially, when bigger training datasets are available. In deep learning applications, discrete data, e.g. words or n-grams in language, or amino acids or nucleotides in bioinformatics, are generally represented as a continuous vector through an embedding matrix. Recently, learning this embedding matrix directly from the data as part of the continuous iteration of the model to optimize the target prediction – a process called ‘end-to-end learning’ – has led to state-of-the-art results in many fields. Although usage of embeddings is well described in the bioinformatics literature, the potential of end-to-end learning for single amino acids, as compared to more classical manually-curated encoding strategies, has not been systematically addressed. To this end, we compared classical encoding matrices, namely one-hot, VHSE8 and BLOSUM62, to end-to-end learning of amino acid embeddings for two different prediction tasks using three widely used architectures, namely recurrent neural networks (RNN), convolutional neural networks (CNN), and the hybrid CNN-RNN. Results: By using different deep learning architectures, we show that end-to-end learning is on par with classical encodings for embeddings of the same dimension even when limited training data is available, and might allow for a reduction in the embedding dimension without performance loss, which is critical when deploying the models to devices with limited computational capacities. We found that the embedding dimension is a major factor in controlling the model performance. Surprisingly, we observed that deep learning models are capable of learning from random vectors of appropriate dimension. Conclusion: Our study shows that end-to-end learning is a flexible and powerful method for amino acid encoding. Further, due to the flexibility of deep learning systems, amino acid encoding schemes should be benchmarked against random vectors of the same dimension to disentangle the information content provided by the encoding scheme from the distinguishability effect provided by the scheme.

Deep-learning↗

NASA NTRS2020

Amino Acid Encoding for Deep Learning Applications

Background: The number of applications of deep learning algorithms in bioinformatics is increasing as they usually achieve superior performance over classical approaches, especially, when bigger training datasets are available. In deep learning applications, discrete data, e.g. words or n-grams in language, or amino acids or nucleotides in bioinformatics, are generally represented as a continuous vector through an embedding matrix. Recently, learning this embedding matrix directly from the data as part of the continuous iteration of the model to optimize the target prediction – a process called ‘end-to-end learning’ – has led to state-ofthe-art results in many fields. Although usage of embeddings is well described in the bioinformatics literature, the potential of end-to-end learning for single amino acids, as compared to more classical manually-curated encoding strategies, has not been systematically addressed. To this end, we compared classical encoding matrices, namely one-hot, VHSE8 and BLOSUM62, to end-to-end learning of amino acid embeddings for two different prediction tasks using three widely used architectures, namely recurrent neural networks (RNN), convolutional neural networks (CNN), and the hybrid CNN-RNN. Results: By using different deep learning architectures, we show that end-to-end learning is on par with classical encodings for embeddings of the same dimension even when limited training data is available, and might allow for a reduction in the embedding dimension without performance loss, which is critical when deploying the models to devices with limited computational capacities. We found that the embedding dimension is a major factor in controlling the model performance. Surprisingly, we observed that deep learning models are capable of learning from random vectors of appropriate dimension. Conclusion: Our study shows that end-to-end learning is a flexible and powerful method for amino acid encoding. Further, due to the flexibility of deep learning systems, amino acid encoding schemes should be benchmarked against random vectors of the same dimension to disentangle the information content provided by the encoding scheme from the distinguishability effect provided by the scheme.

Hesham ElAbd↗

NASA NTRS2005

Engineering Lessons Learned and Systems Engineering Applications

Systems Engineering is fundamental to good engineering, which in turn depends on the integration and application of engineering lessons learned and technical standards. Thus, good Systems Engineering also depends on systems engineering lessons learned from within the aerospace industry being documented and applied. About ten percent of the engineering lessons learned documented in the NASA Lessons Learned Information System are directly related to Systems Engineering. A key issue associated with lessons learned datasets is the communication and incorporation of this information into engineering processes. Systems Engineering has been defined (EINIS-632) as "an interdisciplinary approach encompassing the entire technical effort to evolve and verify an integrated and life-cycle balanced set of system people, product, and process solutions that satisfy customer needs". Designing reliable space-based systems has always been a goal for NASA, and many painful lessons have been learned along the way. One of the continuing functions of a system engineer is to compile development and operations "lessons learned" documents and ensure their integration into future systems development activities. They can produce insights and information for risk identification identification and characterization. on a new project. Lessons learned files from previous projects are especially valuable in risk

Gill, Paul S.↗

DOE OSTI2025

A resolution independent neural operator

The Deep operator network (DeepONet) is a powerful yet simple neural operator architecture that utilizes two deep neural networks to learn mappings between infinite-dimensional function spaces. This architecture is highly flexible, allowing the evaluation of the solution field at any location within the desired domain. However, it imposes a strict constraint on the input space, requiring all input functions to be discretized at the same locations; this limits its practical applications. Here, in this work, we introduce a general framework for operator learning from input–output data with arbitrary number and locations of sensors. This begins by introducing a resolution-independent DeepONet (RI-DeepONet), enabling it to handle input functions that are arbitrarily, but sufficiently finely, discretized. To this end, we propose two dictionary learning algorithms to adaptively learn a set of appropriate continuous basis functions, parameterized as implicit neural representations (INRs), from correlated signals defined on arbitrary point cloud data. These basis functions are then used to project arbitrary input function data as a point cloud onto an embedding space (i.e., a vector space of finite dimensions) with dimensionality equal to the dictionary size, which can be directly used by DeepONet without any architectural changes. In particular, we utilize sinusoidal representation networks (SIRENs) as trainable INR basis functions. The introduced dictionary learning algorithms are then used in a similar way to learn an appropriate dictionary of basis functions for the output function data, which defines a new neural operator architecture referred to as the R esolution I ndependent N eural O perator (RINO). In the RINO, the operator learning task simplifies to learning a mapping from the coefficients of input basis functions to the coefficients of output basis functions. We demonstrate the robustness and applicability of RINO in handling arbitrarily (but sufficiently richly) sampled input and output functions during both training and inference through several numerical examples.

Deep operator network (DeepONet)↗

NASA NTRS2001

Lessons Learned (3 Years of H2O2 Propulsion System Testing Efforts at NASA's John C. Stennis Space Center)

John C. Stennis Space Center continues to support the Propulsion community in an effort to validate High-Test Peroxide as an alternative to existing/future oxidizers. This continued volume of peroxide test/handling activity at Stennis Space Center (SSC) provides numerous opportunities for the SSC team to build upon previously documented 'lessons learned'. SSC shall continue to strive to document their experience and findings as H2O2 issues surface. This paper is intended to capture all significant peroxide issues that we have learned over the last three years. This data (lessons learned) has been formulated from practical handling, usage, storage, operations, and initial development/design of our systems/facility viewpoint. The paper is intended to be an information type tool and limited in technical rational; therefore, presenting the peroxide community with some issues to think about as the continued interest in peroxide evolves and more facilities/hardware are built. These lessons learned are intended to assist industry in mitigating problems and identifying potential pitfalls when dealing with the requirements for handling high-test peroxide.

Taylor, Gary O.↗