Search NASA⌕ Search

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.

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

At least 109 records · Page 6
All sourcesNASA NTRSDOE OSTI
DOE OSTI2022

Field Emission Mitigation in CEBAF SRF Cavities Using Deep Learning

The Continuous Electron Beam Accelerator Facility (CEBAF) operates hundreds of superconducting radio frequency (SRF) cavities in its two main linear accelerators. Field emission can occur when the cavities are set to high operating RF gradients and is an ongoing operational challenge. This is especially true in newer, higher gradient SRF cavities. Field emission results in damage to accelerator hardware, generates high levels of neutron and gamma radiation, and has deleterious effects on CEBAF operations. So, field emission reduction is imperative for the reliable, high gradient operation of CEBAF that is required by experimenters. Here we explore the use of deep learning architectures via multilayer perceptron to simultaneously model radiation measurements at multiple detectors in response to arbitrary gradient distributions. These models are trained on collected data and could be used to minimize the radiation production through gradient redistribution. This work builds on previous efforts in developing machine learning (ML) models, and is able to produce similar model performance as our previous ML model without requiring knowledge of the field emission onset for each cavity.

Ahammed, K.↗
DOE OSTI2022

Field Emission Mitigation in CEBAF SRF Cavities Using Deep Learning

The Continuous Electron Beam Accelerator Facility (CEBAF) operates hundreds of superconducting radio frequency (SRF) cavities in its two main linear accelerators. Field emission can occur when the cavities are set to high operating RF gradients and is an ongoing operational challenge. This is especially true in newer, higher gradient SRF cavities. Field emission results in damage to accelerator hardware, generates high levels of neutron and gamma radiation, and has deleterious effects on CEBAF operations. So, field emission reduction is imperative for the reliable, high gradient operation of CEBAF that is required by experimenters. Here we explore the use of deep learning architectures via multilayer perceptron to simultaneously model radiation measurements at multiple detectors in response to arbitrary gradient distributions. These models are trained on collected data and could be used to minimize the radiation production through gradient redistribution. This work builds on previous efforts in developing machine learning (ML) models, and is able to produce similar model performance as our previous ML model without requiring knowledge of the field emission onset for each cavity.

Ahammed, K.↗
NASA NTRSDate not supplied

Towards Autonomous Lunar Resource Excavation via Reinforcement Learning

To continue on a sustainable and flexible path, NASA needs to address the challenge of collecting and moving large amounts of regolith at the destination. NASA’s Regolith Advanced Surface Systems Operations Robot (RASSOR) is principally designed to mine and deliver regolith for In-Situ Resource Utilization (ISRU) processing. RASSOR’s design enables it to efficiently collect and deposit regolith, return collected material for processing, and myriad related ISRU activities. To reliably perform these operations on the lunar surface, RASSOR software and sensory systems need to be robust and maximize the information extracted from a reduced sensor payload. Herein, we present preliminary findings from the Intelligent Capabilities Enhanced RASSOR project. We created reduced-order simulation environments to develop autonomous trenching controllers via reinforcement learning and prototype state estimation architectures. The goal of reinforcement learning is for an agent to learn a policy (task strategy) through interactions with an environment. When the agent performs an action, a change occurs in environment state and a numerical reward is received which informs the agent whether the action performed was good or not. Since reinforcement learning algorithms learn through trial-and-error, a simulation is a desirable first environment for development and learning. We developed two simulations, the first is a 2D excavation simulation developed to facilitate parameter selection, and a 3D simulation developed using a game physics engine, to simulate simplified soil interactions and increase the fidelity of the dynamic models of the robotic agents. The development of this 3D simulation has enabled the training of additional sensing capabilities and research both at the granular mechanics and operations levels. We experimented with various virtual sensor payloads to identify a combination that enabled efficient excavation operation and learning. Our reward function is based on how much material is excavated per step. A penalty is also received for leaving the dig site and to smooth the acceleration of the drum arms. We implemented pseudo time-of-flight sensors to report distance from each drum to ground and the height above ground which was found to be more efficient than existing solutions. Our findings suggest that reinforcement learning for autonomous operations has learned viable trenching strategies within 3000 training episodes in our simplified 2D environment and helped identify desirable sensing capabilities, arrangements, and considerations such as the positioning of time-of-flight sensors. Future work includes expanding our simulation to more complex environments and scenarios, and transfer learning from simulation to RASSOR 2.0 hardware for deployment in the Regolith Test Bin at NASA's Kennedy Space Center.

rassor↗
DOE OSTI2022

Battery calendar aging and machine learning

Batteries continue to attract immense attention due to the critical need for new energy technologies that can help humankind transition to a net-zero future. To meet global goals for implementation of new batteries, the research community needs to more coherently consider how to go about the process of innovation, development, and validation of new technologies. Current practices for bringing a new battery technology from the benchtop to deployment can be very time consuming and complex, given that batteries can have distinct aging pathways based on both how they are cycled and their storage conditions. In nearly all use cases, batteries experience periods of active cycling and also extended periods of rest. Thus, calendar time often occurs when a battery is close to fully charged. Currently, the combined cycle and calendar life aspects receive inconsistent attention during most stages of research and development. For batteries to fulfill the critical role envisioned to meet global energy demands, greater uniformity in practice is needed to alleviate potential delays caused by the inconsistent acquisition of aging data. In this study, we present our thoughts and recommendations on the need to further advance commonality and availability of calendar aging data for emerging high-energy and long-duration batteries. While there is a broad need to adopt best practices for both calendar and cycle life assessment, a push to greater uniformity in approach and reporting is sorely needed to allow calendar life predictions to harness many of the advances that have recently been seen for cycle life prediction using machine learning (ML) and other techniques.

25 ENERGY STORAGE↗
DOE OSTI2022

Biological underpinnings for lifelong learning machines

Biological organisms learn from interactions with their environment throughout their lifetime. For artificial systems to successfully act and adapt in the real world, it is desirable to similarly be able to learn on a continual basis. This challenge is known as lifelong learning, and remains to a large extent unsolved. In this Perspective article, we identify a set of key capabilities that artificial systems will need to achieve lifelong learning. We describe a number of biological mechanisms, both neuronal and non-neuronal, that help explain how organisms solve these challenges, and present examples of biologically inspired models and biologically plausible mechanisms that have been applied to artificial systems in the quest towards development of lifelong learning machines. We discuss opportunities to further our understanding and advance the state of the art in lifelong learning, aiming to bridge the gap between natural and artificial intelligence. It is an outstanding challenge to develop intelligent machines that can learn continually from interactions with their environment, throughout their lifetime. Kudithipudi et al. review neuronal and non-neuronal processes in organisms that address this challenge and discuss pathways to developing biologically inspired approaches for lifelong learning machines.

Kudithipudi, Dhireesha↗
DOE OSTI2026

Decoding α-MoC 1− x Nanoparticle Formation in Continuous Flow via Machine Learning

Molybdenum carbide nanoparticles (α-MoC 1−x NPs) are promising catalysts that offer noble-metal-like performance at lower cost. We report a mild continuous-flow synthesis of α-MoC 1−x NPs from Mo(CO) 6 , coupled with in-line spectroscopic monitoring and machine learning (ML)-based analysis to quantify precursor conversion and product formation in real time. A multilayer perceptron ML model was found to accurately deconvolute complex, nonlinear spectral patterns, enabling identification of a two-step reaction pathway, involving precursor conversion to an amorphous intermediate followed by intraparticle crystallization to α-MoC 1−x NPs, with the first step being rate limiting. Ex situ small angle X-ray scattering (SAXS) and X-ray diffraction (XRD) validation confirm the predicted concentration profiles and crystallization behavior. This integrated approach showcases how ML can empower insights into NP nucleation and growth, paving the way for self-driving, flow-based platforms for NP synthesis.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2024

Newton-Raphson AC Power Flow Convergence Based on Deep Learning Initialization and Homotopy Continuation

Power flow forms the basis of many power system studies. With the increased penetration of renewable energy, grid planners tend to perform multiple power flow simulations under various operating conditions and not just selected snapshots at peak or light load conditions. Getting a converged AC power flow (ACPF) case remains a significant challenge for grid planners especially in large power grid networks. This paper proposes a two-stage approach to improve Newton-Raphson ACPF convergence and was applied to a 6102 bus Electric Reliability Council of Texas (ERCOT) system. The first stage utilizes a deep learning-based initializer with data re-training. Here a deep neural network (DNN) initializer is developed to provide better initial voltage magnitude and angle guesses to aid in power flow convergence. This is because Newton-Raphson ACPF is quite sensitive to the initial conditions and bad initialization could lead to divergence. The DNN initializer includes a data re-training framework that improves the initializer's performance when faced with limited training data. The DNN initializer successfully solved 3,285 cases out of 3,899 non-converging dispatch and performed better than random forest and DC power flow initialization methods. ACPF cases not solved in this first stage are then passed through a hot-starting algorithm based on homotopy continuation with switched shunt control. The hot-starting algorithm successfully converged 416 cases out of the remaining 614 non-converging ACPF dispatch. In conclusion, the combined two-stage approach achieved a 94.9% success rate, by converging a total of 3,701 cases out of the initial 3,899 unsolved cases.

Deep learning↗
DOE OSTI2012

Savannah River Site, Liquid Waste Program, Savannah River Remediation American Recovery and Reinvestment Act Benefits and Lessons Learned - 12559

Utilizing funding provided by the American Recovery and Reinvestment Act (ARRA), the Liquid Waste Program at Savannah River site successfully executed forty-one design, procurement, construction, and operating activities in the period from September 2009 through December 2011. Project Management of the program included noteworthy practices involving safety, integrated project teams, communication, and cost, schedule and risk management. Significant upgrades to plant capacity, progress toward waste tank closure and procurement of needed infrastructure were accomplished. Over 1.5 million hours were worked without a single lost work day case. Lessons Learned were continually identified and applied to enhance the program. Investment of Recovery Act monies into the Liquid Waste Program has ensured continued success in the disposition of radioactive wastes and the closure of high level waste tanks at SRS. The funding of a portion of the Liquid Waste Program at SRS by ARRA was a major success. Significant upgrades to plant capacity, progress toward waste tank closure and procurement of needed infrastructure was accomplished. Integrated Project Teams ensured quality products and services were provided to the Operations customers. Over 1.5 million hours were worked without a single lost work day case. Lessons Learned were continually reviewed and reapplied to enhance the program. Investment of Recovery Act monies into the Liquid Waste Program has ensured continued success in the disposition of radioactive wastes and the closure of high level waste tanks at SRS. (authors)

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗
DOE OSTI2018

HyperSpace: Distributed Bayesian Hyperparameter Optimization

As machine learning models continue to increase in complexity, so does the potential number of free model parameters commonly known as hyperparameters. While there has been considerable progress toward finding optimal configurations of these hyperparameters, many optimization procedures are treated as black boxes. We believe optimization methods should not only return a set of optimized hyperparameters, but also give insight into the effects of model hyperparameter settings. To this end, we present HyperSpace, a parallel implementation of Bayesian sequential model-based optimization. HyperSpace leverages high performance computing (HPC) resources to better understand unknown, potentially non-convex hyperparameter search spaces. We show that it is possible to learn the dependencies between model hyperparameters through the optimization process. By partitioning large search spaces and running many optimization procedures in parallel, we also show that it is possible to discover families of good hyperparameter settings over a variety of models including unsupervised clustering, regression, and classification tasks.

Computer Science↗
DOE OSTI2021

Utilizing ensemble learning for performance and power modeling and improvement of parallel cancer deep learning CANDLE benchmarks

Abstract Machine learning (ML) continues to grow in importance across nearly all domains in modeling to learn from data. Often a tradeoff exists between a model's ability to minimize bias and variance. In this article, we utilize ensemble learning to combine linear, nonlinear, and tree‐/rule‐based ML methods to cope with the bias‐variance tradeoff and result in more accurate models. We use the datasets collected for two parallel cancer deep learning CANDLE benchmarks, NT3 and P1B2, to build performance and power models based on hardware performance counters using single‐object and multiple‐objects ensemble learning to identify the most important counters for improvement on the Cray XC40 Theta at Argonne National Laboratory. Based on the insights from these models, we improve the performance and energy of P1B2 and NT3 by optimizing the deep learning environments TensorFlow, Keras, Horovod, and Python under the huge page size of 8 MB. Experimental results show that ensemble learning not only produces more accurate models but also provides more robust performance counter ranking. We achieve up to 61.15% performance improvement and up to 62.58% energy saving for P1B2 and up to 55.81% performance improvement and up to 52.60% energy saving for NT3 on up to 24,576 cores.

Wu, Xingfu↗
DOE OSTI2013

Multiple transient memories in sheared suspensions: Robustness, structure, and routes to plasticity

Multiple transient memories, first discovered in charge-density-wave conductors, are a remarkable and initially counterintuitive example of how a system can store information about its driving. In this class of memories, a system can learn multiple driving inputs, nearly all of which are eventually forgotten despite their continual input. If sufficient noise is present, the system regains plasticity so that it can continue to learn new memories indefinitely. Recently, Keim and Nagel [Phys. Rev. Lett. 107, 010603 (2011)] showed how multiple transient memories could be generalized to a generic driven disordered system with noise, giving as an example simulations of a simple model of a sheared non-Brownian suspension. Here, we further explore simulation models of suspensions under cyclic shear, focusing on three main themes: robustness, structure, and overdriving. We show that multiple transient memories are a robust feature independent of many details of the model. The steady-state spatial distribution of the particles is sensitive to the driving algorithm; nonetheless, the memory formation is independent of such a change in particle correlations. Finally, we demonstrate that overdriving provides another means for controlling memory formation and retention.

42 ENGINEERING↗
DOE OSTI2021

Latent map Gaussian processes for mixed variable metamodeling

Gaussian processes (GPs) are ubiquitously used in sciences and engineering as metamodels. Standard GPs, however, can only handle numerical or quantitative variables. Here we introduce latent map Gaussian processes (LMGPs) that inherit the attractive properties of GPs and are also applicable to mixed data which have both quantitative and qualitative inputs. The core idea behind LMGPs is to learn a continuous, low-dimensional latent space or manifold which encodes all qualitative inputs. To learn this manifold, we first assign a unique prior vector representation to each combination of qualitative inputs. We then use a low-rank linear map to project these priors on a manifold that characterizes the posterior representations. As the posteriors are quantitative, they can be directly used in any standard correlation function such as the Gaussian or Matern. Hence, the optimal map and the corresponding manifold, along with other hyperparameters of the correlation function, can be systematically learned via maximum likelihood estimation. Through a wide range of analytic and real-world examples, we demonstrate the advantages of LMGPs over state-of-the-art methods in terms of accuracy and versatility. In particular, we show that LMGPs can handle variable-length inputs, have an explainable neural network interpretation, and provide insights into how qualitative inputs affect the response or interact with each other. We also employ LMGPs in Bayesian optimization and illustrate that they can discover optimal compound compositions more efficiently than conventional methods that convert compositions to qualitative variables via manual featurization.

42 ENGINEERING↗
NASA NTRS2005

Spatial Predictive Modeling and Remote Sensing of Land Use Change in the Chesapeake Bay Watershed

This project was focused on modeling the processes by which increasing demand for developed land uses, brought about by changes in the regional economy and the socio-demographics of the region, are translated into a changing spatial pattern of land use. Our study focused on a portion of the Chesapeake Bay Watershed where the spatial patterns of sprawl represent a set of conditions generally prevalent in much of the U.S. Working in the region permitted us access to (i) a time-series of multi-scale and multi-temporal (including historical) satellite imagery and (ii) an established network of collaborating partners and agencies willing to share resources and to utilize developed techniques and model results. In addition, a unique parcel-level tax assessment database and linked parcel boundary maps exists for two counties in the Maryland portion of this region that made it possible to establish a historical cross-section time-series database of parcel level development decisions. Scenario analyses of future land use dynamics provided critical quantitative insight into the impact of alternative land management and policy decisions. These also have been specifically aimed at addressing growth control policies aimed at curbing exurban (sprawl) development. Our initial technical approach included three components: (i) spatial econometric modeling of the development decision, (ii) remote sensing of suburban change and residential land use density, including comparisons of past change from Landsat analyses and more traditional sources, and (iii) linkages between the two through variable initialization and supplementation of parcel level data. To these we added a fourth component, (iv) cellular automata modeling of urbanization, which proved to be a valuable addition to the project. This project has generated both remote sensing and spatially explicit socio-economic data to estimate and calibrate the parameters for two different types of land use change models and has undertaken analyses of these models. One (the CA model) is driven largely by observations on past patterns of land use change, while the other (the EC model) is driven by mechanisms of the land use change decision at the parcel level. Our project may be the first serious attempt at developing both types of models for the same area, using as much common data as possible. We have identified the strengths and weaknesses of the two approaches and plan to continue to revise each model in the light of new data and new lessons learned through continued collaboration. Questions, approaches, findings, publication and presentation lists concerning the research are also presented.

Goetz, Scott J.↗
DOE OSTI2022

Kernel learning backward SDE filter for data assimilation

In this paper, we develop a kernel learning backward SDE filter method to estimate the state of a stochastic dynamical system based on its partial noisy observations. A system of forward backward stochastic differential equations is used to propagate the state of the target dynamical model, and Bayesian inference is applied to incorporate the observational information. Further, to characterize the dynamical model in the entire state space, we introduce a kernel learning method to learn a continuous global approximation for the conditional probability density function of the target state by using discrete approximated density values as training data. Numerical experiments demonstrate that the kernel learning backward SDE is highly effective.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2016

Continous Representation Learning via User Feedback

Representation learning is a deep-learning based technique for extracting features from data for the purpose of machine learning. This requires a large amount of data, on order tens of thousands to millions of samples, to properly teach the deep neural network. This a system for continuous representation learning, where the system may be improved with a small number of additional samples (order 10-100). The unique characteristics of this invention include a human-computer feedback component, where assess the quality of the current representation and then provides a better representation to the system. The system then mixes the new data with old training examples to avoid overfitting and improve overall performance of the system. The model can be exported and shared with other users, and it may be applied to additional images the system hasn't seen before.

Source record↗