SEARCH · Search NASA

Results for “Learning with errors”

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 181 records · Page 10

All sources NASA NTRS DOE OSTI

NASA NTRS2017

NeMO-Net: The Neural Multi-Modal Observation and Training Network for Global Coral Reef Assessment

In the past decade, coral reefs worldwide have experienced unprecedented stresses due to climate change, ocean acidification, and anthropomorphic pressures, instigating massive bleaching and die-off of these fragile and diverse ecosystems. Furthermore, remote sensing of these shallow marine habitats is hindered by ocean wave distortion, refraction and optical attenuation, leading invariably to data products that are often of low resolution and signal-to-noise (SNR) ratio. However, recent advances in UAV and Fluid Lensing technology have allowed us to capture multispectral 3D imagery of these systems at sub-cm scales from above the water surface, giving us an unprecedented view of their growth and decay. Exploiting the fine-scaled features of these datasets, machine learning methods such as MAP, PCA, and SVM can not only accurately classify the living cover and morphology of these reef systems (below 8 percent error), but are also able to map the spectral space between airborne and satellite imagery, augmenting and improving the classification accuracy of previously low-resolution datasets. We are currently implementing NeMO-Net, the first open-source deep convolutional neural network (CNN) and interactive active learning and training software to accurately assess the present and past dynamics of coral reef ecosystems through determination of percent living cover and morphology. NeMO-Net will be built upon the QGIS platform to ingest UAV, airborne and satellite datasets from various sources and sensor capabilities, and through data-fusion determine the coral reef ecosystem makeup globally at unprecedented spatial and temporal scales. To achieve this, we will exploit virtual data augmentation, the use of semi-supervised learning, and active learning through a tablet platform allowing for users to manually train uncertain or difficult to classify datasets. The project will make use of Pythons extensive libraries for machine learning, as well as extending integration to GPU and High-End Computing Capability (HECC) on the Pleiades supercomputing cluster, located at NASA Ames. The project is being supported by NASAs Earth Science Technology Office (ESTO) Advanced Information Systems Technology (AIST-16) Program.

NeMO-Net↗

NASA NTRS1997

Lessons Learned from the Clementine Mission

According to BMDO, the Clementine mission achieved many of its technology objectives during its flight to the Moon in early 1994 but, because of a software error, was unable to test the autonomous tracking of a cold target. The preliminary analyses of the returned lunar data suggest that valuable scientific measurements were made on several important topics but that COMPLEX's highest-priority objectives for lunar science were not achieved. This is not surprising given that the rationale for Clementine was technological rather than scientific. COMPLEX lists below a few of the lessons that may be learned from Clementine. Although the Clementine mission was not conceived as a NASA science mission exactly like those planned for the Discovery program, many operational aspects of the two are similar. It is therefore worthwhile to understand the strengths and faults of the Clementine approach. Some elements of the Clementine operation that led to the mission's success include the following: (1) The mission's achievements were the responsibility of a single organization and its manager, which made that organization and that individual accountable for the final outcome; (2) The sponsor adopted a hands-off approach and set a minimum number of reviews (three); (3) The sponsor accepted a reasonable amount of risk and allowed the project team to make the trade-offs necessary to minimize the mission's risks while still accomplishing all its primary objectives; and (4) The development schedule was brief and the agreed-on funding (and funding profile) was adhered to. Among the operational shortcomings of Clementine were the following: (1) An overly ambitious schedule and a slightly lean budget (meaning insufficient time for software development and testing, and leading ultimately to human exhaustion); and (2) No support for data calibration, reduction, and analysis. The principal lesson to be learned in this category is that any benefits from the constructive application of higher risk for lower cost and faster schedule will be lost if the schedule does not allow adequate time for the development of all essential systems or makes no allowance for human frailties. Another lesson to be drawn is that despite its limitations, if judged strictly as a science mission, Clementine attested that significant scientific information can be gathered during a technology-demonstration mission. In the current era of limited funds, when science missions will be infrequent, the opportunity to fly scientific instruments aboard missions whose objectives might be other than science must be seized and, indeed, encouraged. During such opportunities it would be inexcusable to do second-class science. Thus the scientific community must be actively involved in such projects from their initiation.

Source record↗

NASA NTRS1994

Application of artificial neural networks in hydrological modeling: A case study of runoff simulation of a Himalayan glacier basin

The simulation of runoff from a Himalayan Glacier basin using an Artificial Neural Network (ANN) is presented. The performance of the ANN model is found to be superior to the Energy Balance Model and the Multiple Regression model. The RMS Error is used as the figure of merit for judging the performance of the three models, and the RMS Error for the ANN model is the latest of the three models. The ANN is faster in learning and exhibits excellent system generalization characteristics.

Buch, A. M.↗

NASA NTRS1998

Modeling learning in brain stem and cerebellar sites responsible for VOR plasticity

A simple model of vestibuloocular reflex (VOR) function was used to analyze several hypotheses currently held concerning the characteristics of VOR plasticity. The network included a direct vestibular pathway and an indirect path via the cerebellum. An optimization analysis of this model suggests that regulation of brain stem sites is critical for the proper modification of VOR gain. A more physiologically plausible learning rule was also applied to this network. Analysis of these simulation results suggests that the preferred error correction signal controlling gain modification of the VOR is the direct output of the accessory optic system (AOS) to the vestibular nuclei vs. a signal relayed through the cerebellum via floccular Purkinje cells. The potential anatomical and physiological basis for this conclusion is discussed, in relation to our current understanding of the latency of the adapted VOR response.

NASA Discipline Neuroscience↗

NASA NTRS2019

Exploring Beyond Earth's Atmosphere with Human-Machine Teams

NASA's highly successful Kepler Mission has revolutionized our understanding of the Galaxy. We now know that planets, even Earth-size planets in the habitable zone, are common. With the end of the Kepler Mission we now look to the future with the Transiting Exoplanet Survey Satellite (TESS) which will discover thousands of exoplanets in orbit around the brightest stars in the sky. In a two-year survey, TESS will perform an all-sky search of more than 200,000 stars for temporary drops in brightness caused by planetary transits. With Kepler and TESS, humanity is finally at the verge of studying the masses, sizes, densities, orbits, and atmospheres of a large cohort of small planets, including a sample of rocky worlds in the habitable zones of their host stars which may prove to host life. The massive data sets generated by Kepler and TESS must be meticulously combed for the weakest planetary signals every month. While a daunting and error-prone task for humans, this is an exciting opportunity for the breakthroughs recently seen in machine learning. Specifically, traditional methods for identifying planet transits require extensive data processing pipelines followed by extensive human vetting. This manual process risks loss of information due to the data processing and to inconsistency and biases due to individual human vetters. The latest advancements in machine learning will allow an objective classifier to minimize the losses of information and greatly lessen the burden on the human vetters, in addition to providing assessment of quality and score to each planet candidate, freeing the humans to concentrate on border cases and other more interesting investigations.

Smith, Jeffrey C.↗

NASA NTRSDate not supplied

Machine Learning based Aircraft Performance Model Estimation for Trajectory Prediction

The accurate prediction of aircraft trajectory by ground-based decision support tools is a critical component of air traffic management in the US National Airspace System (NAS). Accurate predictions of where the aircraft will be in the future or when they will arrive at specific locations (e.g., fixes) is a key enabler for sequencing and efficient arrival management of flights. Traditional physics based aircraft trajectory prediction relies on a simplified point-mass total energy model whose parameters are referred to as Aircraft Performance Model (APM) parameters. Even though the performance coefficients and weight of an aircraft are a vital part of the aircraft performance model’s predictions and accuracy, these coefficients are proprietary in nature and therefore, unavailable to decision-support tools. Current approaches freeze some coefficients to default base of aircraft data (BADA) values and optimize others. However, the APM parameters are highly coupled by the flight dynamics and prioritizing one parameter over others leads to bias and skewed predictions. To alleviate this problem, we provide a combined optimization framework to predict all the critical (thrust, drag and weight) APM parameters. This paper is focused on training Machine Learning (ML) models that map historical flights to optimized APM parameters that provide the best fit (in terms of prediction error). Our dataset obtained from NASA’s Sherlock data warehouse is comprised of thousands of historical flights and includes weather and track data collected from 2019. Using different subsets of relevant features (e.g., aircraft type), we trained several ML models to estimate the aircraft’s take off weight, drag polar coefficients (both parasitic and lift induced), and thrust settings (multiplier applied to the maximum engine thrust). The chosen flights are from three of the most common aircraft types (B738, B737, and A320) arriving at four airports (LAX, DEN, MSP, and DFW). Our ML approach is comprised of two different solutions: 1- using a subset of features that are known prior to the flight departure and do not change during flight (such as engine type, current temperature at departure & destination airports, aircraft type) and 2 - using a subset of temporal features of the flight trajectory (such as cruise altitude, Mach, airspeed, and rate of climb) in addition to the pre-departure features from the first solution. The labels or target variables are the APM parameters that were obtained by an optimized ordinary differential equations (ODE) fitting process (applied to individual flights). The ODE-fitting is very time intensive and is therefore performed offline. Thus, training an ML model to learn the relationship between the flight features and ODE-generated labels enables faster estimation of the APM parameters and is therefore amenable to real-time prediction. Various ML models including linear regression, random forest, XGBoost, and neural network were trained, and the results are compared. After model validation and hyperparameter-tuning, we observed that the Random Forest model outperformed the other three models by the overall mean square error (MSE) of 2% for the first solution and 1.5% for the second solution. Finally, the ML-derived parameters are compared against default BADA APM parameters using NASA’s Autonomy Development toolkit (ADK) simulation software. The simulation results for one of each aircraft type is shown and discussed.

Aida Sharif Rohani↗

NASA NTRS2020

Tuning Neural Network Models for Improved Prediction of Boundary Layer Transition

Boundary layer transition can strongly impact flight vehicle performance as it influences surface skin friction and aerodynamic heating, making accurate transition prediction a key to designing next generation aircraft. Artificial neural networks (ANNs) have shown promise toward predicting laminar-turbulent transition based on linear stability correlations. The computational efficiency of ANNs and the substantially reduced user involvement in relation to direct computations based on the linear stability theory (LST) makes them an attractive methodology for integrating the LST based correlations in computational fluid dynamics codes. Tollmien-Schlichting (TS) waves correspond to the dominant transition mechanism in 2D or weakly 3D subsonic boundary layers, such as those encountered in general aviation applications. Improvements to neural network model accuracy in predicting the amplification rates of TS instability waves have been investigated by leveraging recent machine learning developments in conjunction with surrogate optimization techniques and via suitable augmentation of the data used to train the networks. The optimized models trained on the modified dataset reduced the average transition location errors on different airfoils at several flow conditions by 51% of the original manually-tuned network’s errors on the same flow cases. The actual transition locations were derived from the Langley Stability and Transition Analysis Code (LASTRAC).

Machine Learning↗

NASA NTRS1993

An expert system shell for inferring vegetation characteristics: Changes to the historical cover type database (Task F)

All the options in the NASA VEGetation Workbench (VEG) make use of a database of historical cover types. This database contains results from experiments by scientists on a wide variety of different cover types. The learning system uses the database to provide positive and negative training examples of classes that enable it to learn distinguishing features between classes of vegetation. All the other VEG options use the database to estimate the error bounds involved in the results obtained when various analysis techniques are applied to the sample of cover type data that is being studied. In the previous version of VEG, the historical cover type database was stored as part of the VEG knowledge base. This database was removed from the knowledge base. It is now stored as a series of flat files that are external to VEG. An interface between VEG and these files was provided. The interface allows the user to select which files of historical data to use. The files are then read, and the data are stored in Knowledge Engineering Environment (KEE) units using the same organization of units as in the previous version of VEG. The interface also allows the user to delete some or all of the historical database units from VEG and load new historical data from a file. This report summarizes the use of the historical cover type database in VEG. It then describes the new interface to the files containing the historical data. It describes minor changes that were made to VEG to enable the externally stored database to be used. Test runs to test the operation of the new interface and also to test the operation of VEG using historical data loaded from external files are described. Task F was completed. A Sun cartridge tape containing the KEE and Common Lisp code for the new interface and the modified version of the VEG knowledge base was delivered to the NASA GSFC technical representative.

Source record↗

NASA NTRS2002

Resource Selection Using Execution and Queue Wait Time Predictions

Computational grids provide users with many possible places to execute their applications. We wish to help users select where to run their applications by providing predictions of the execution times of applications on space shared parallel computers and predictions of when scheduling systems for such parallel computers will start applications. Our predictions are based on instance based learning techniques and simulations of scheduling algorithms. We find that our execution time prediction techniques have an average error of 37 percent of the execution times for trace data recorded from SGI Origins at NASA Ames Research Center and that this error is 67 percent lower than the error of user estimates. We also find that the error when predicting how long applications will wait in scheduling queues is 95 percent of mean queue wait times when using our execution time predictions and this is 57 percent lower than if we use user execution time estimates.

Warren, Smith↗

NASA NTRS2002

Resource Selection Using Execution and Queue Wait Time Predictions

Smith, Warren↗

NASA NTRS2006

The NASA F-15 Intelligent Flight Control Systems: Generation II

The Second Generation (Gen II) control system for the F-15 Intelligent Flight Control System (IFCS) program implements direct adaptive neural networks to demonstrate robust tolerance to faults and failures. The direct adaptive tracking controller integrates learning neural networks (NNs) with a dynamic inversion control law. The term direct adaptive is used because the error between the reference model and the aircraft response is being compensated or directly adapted to minimize error without regard to knowing the cause of the error. No parameter estimation is needed for this direct adaptive control system. In the Gen II design, the feedback errors are regulated with a proportional-plus-integral (PI) compensator. This basic compensator is augmented with an online NN that changes the system gains via an error-based adaptation law to improve aircraft performance at all times, including normal flight, system failures, mispredicted behavior, or changes in behavior resulting from damage.

Buschbacher, Mark↗

NASA NTRS1992

Learning a trajectory using adjoint functions and teacher forcing

A new methodology for faster supervised temporal learning in nonlinear neural networks is presented which builds upon the concept of adjoint operators to allow fast computation of the gradients of an error functional with respect to all parameters of the neural architecture, and exploits the concept of teacher forcing to incorporate information on the desired output into the activation dynamics. The importance of the initial or final time conditions for the adjoint equations is discussed. A new algorithm is presented in which the adjoint equations are solved simultaneously (i.e., forward in time) with the activation dynamics of the neural network. We also indicate how teacher forcing can be modulated in time as learning proceeds. The results obtained show that the learning time is reduced by one to two orders of magnitude with respect to previously published results, while trajectory tracking is significantly improved. The proposed methodology makes hardware implementation of temporal learning attractive for real-time applications.

Toomarian, Nikzad B.↗

NASA NTRS1979

Learning with imperfectly labeled patterns

The problem of learning in pattern recognition using imperfectly labeled patterns is considered. The performance of the Bayes and nearest neighbor classifiers with imperfect labels is discussed using a probabilistic model for the mislabeling of the training patterns. Schemes for training the classifier using both parametric and non parametric techniques are presented. Methods for the correction of imperfect labels were developed. To gain an understanding of the learning process, expressions are derived for success probability as a function of training time for a one dimensional increment error correction classifier with imperfect labels. Feature selection with imperfectly labeled patterns is described.

Chittineni, C. B.↗

NASA NTRS2022

Run Time Assurance for Electric Vertical Takeoff and Landing Aircraft

NASA is conducting research to demonstrate and evaluate the application of Run Time Assurance (RTA) as a means to assure safety in Electric Vertical Takeoff and Landing (eVTOL) aircraft with highly automated or autonomous flight capability supervised by a single onboard pilot. The work described in this report demonstrates an application of RTA and examines the implications for design and analysis of aircraft functions and systems; aircraft safety hazards; safety assurance; development assurance; and pilot tasks and performance. This research effort also seeks to assess the efficacy of the combined application of traditional Functional Hazard Analysis (FHA) and the more modern System Theoretic Process Analysis (STPA) techniques to perform hazard analyses on aircraft with complex automated and autonomous systems and an onboard pilot. During the research effort we developed architectural designs of two alternate eVTOL aircraft, generally following the process characterized in the SAE standards ARP4754 and ARP4761. The design has focused on the control architectures of these aircraft, which are identical except that one incorporates RTA techniques to reduce the criticality of some key software components. Artifacts of this process include a taxonomy of aircraft-level functions, aircraft-level architecture diagrams, aircraft-level functional hazard assessments (AFHA), function allocations onto aircraft systems and subsystems, functional block diagrams for a select set of control-related functions, and system-level functional hazard assessments (SFHA) for those functions. This project has highlighted the notion that DAL D is something of a sweet spot for low-confidence controllers in an RTA-based design. Among the many activities described in DO-178C, the activities related to requirement verifiability, algorithmic accuracy, and test coverage can be the most challenging for the kinds of advanced control techniques that may be desirable in novel UAM designs, such as adaptive control, machine-learning, artificial intelligence, numerical search, and Monte Carlo based algorithms. Moreover, the standard requires that development teams demonstrate that errors leading to unacceptable failure conditions have been removed from the software. The RTA architecture, which cordons off the low-confidence function, makes it much easier to show this for these kinds of algorithms. With regard to the use of STPA and FHA as complementary hazard analysis techniques, our research effort led us to the conclusion that STPA should be used to derive requirements for hardware and software systems and/or components. Also, STPA is a natural complement to other processes in ARP4754A involving design studies and iteration.

Run-time assurance↗

NASA NTRS2023

Easy, Scalable Subsetting of GEDI Point Clouds

The GEDI Subsetter, a Python tool developed for NASA’s Multi-mission Algorithm and Analysis Platform (MAAP), optimizes the accessibility and visualization of GEDI point clouds by enabling users to efficiently subset data in a convenient, scalable manner. Complex science data often requires users to learn new software skills and handle many large files. Handling and cleaning large data sets is tedious and error-prone. These challenges significantly impede analysis. One of the goals of NASA's MAAP is to provide a platform that lowers the barrier to conducting research and analysis at scale. When a group of MAAP users wanted to conduct above-ground biomass estimation using GEDI data, we found that their existing workflow for leveraging GEDI data suffered from the barriers mentioned above. Furthermore, their workflow did not scale easily beyond a small number of granules. We found that existing tools related to GEDI data retrieval and subsetting were too limiting, so the GEDI Subsetter was written to support MAAP users’ needs. Being able to run many subsetting jobs simultaneously in the MAAP, and parallelizing the code itself, has led to significant speed improvements in obtaining relevant data, reducing subsetting time from hours to minutes. MAAP users can now more quickly and easily obtain only the data relevant to their research, by choosing which GEDI collection they want to work with (L1A, L2A, L2B, or L4A), and how they want to subset it, by specifying an area of interest, a temporal range, and relevant attributes. This has significantly reduced the feedback loop for users, allowing them to much more quickly subset GEDI data and begin their analysis. Although the GEDI Subsetter originally targeted users of the MAAP, it is generalized such that it can also be used outside of the MAAP and includes a command-line interface for convenience. Furthermore, with minor modifications, it should be possible to use it with non-GEDI data as the general pattern should be applicable to other sparse/track-based sensors.

Charles Daniels↗

NASA NTRS2023

Machine Learning to Increase the Quality and Repeatability of 3D Printing - Workflow

The imprecise nature of three-dimensional (3D) printing limits the technology’s use beyond prototyping. For production of end-use parts, such as those for aerospace applications, improvements are needed to enhance quality and repeatability. Much of the difficulty in obtaining high quality printed parts lies in finding optimum printing parameters. Currently, this requires trial and error performed by an expert. Finding the optimum printing parameters is also obfuscated by the variation in optimum parameters throughout the part due to part geometry and printer effects. To allow for locally optimized printing parameters, one can envision a machine learning algorithm that could take in an object, predict the best printing parameters, and communicate these parameters to a printer. With this scenario in mind, we developed a tool that can predict and implement locally optimized printing parameters in 3D printing. This tool consists of elements designed to detect errors in a printed part, predict the probability of local flaws occurring at each point in the part, and select the optimal local parameters for the highest quality part given hardware limitations. The results of this work were highlighted in Advanced Materials Technologies. In this paper, we will discuss in greater depth the workflow and algorithms involved with this tool that were not detailed in the journal publication.

additive manufacturing↗

NASA NTRS1985

Test and evaluation of the 2.4-micron photorefractor ocular screening system

An improved 2.4-m photorefractor ocular screening system was tested and evaluated. The photorefractor system works on the principal of obtaining a colored photograph of both human eyes; and, by analysis of the retinal reflex images, certain ocular defects can be detected such a refractive error, strabismus, and lens obstructions. The 2.4-m photorefractory system uses a 35-mm camera with a telephoto lens and an electronic flash attachment. Retinal reflex images obtained from the new 2.4-m system are significantly improved over earlier systems in image quality. Other features were also improved, notably portability and reduction in mass. A total of 706 school age children were photorefracted, 211 learning disabled and 495 middle school students. The total students having abnormal retinal reflexes were 156 or 22 percent, and 133 or 85 percent of the abnormal had refractive error indicated. Ophthalmological examination was performed on 60 of these students and refractive error was verified in 57 or 95 percent of those examined. The new 2.4-m system has a NASA patent pending and is authorized by the FDA. It provides a reliable means of rapidly screening the eyes of children and young adults for vision problems. It is especially useful for infants and other non-communicative children who cannot be screened by the more conventional methods such as the familiar E chart.

Richardson, J. R.↗

NASA NTRS1981

Comparative analysis of techniques for evaluating the effectiveness of aircraft computing systems

Performability analysis is a technique developed for evaluating the effectiveness of fault-tolerant computing systems in multiphase missions. Performability was evaluated for its accuracy, practical usefulness, and relative cost. The evaluation was performed by applying performability and the fault tree method to a set of sample problems ranging from simple to moderately complex. The problems involved as many as five outcomes, two to five mission phases, permanent faults, and some functional dependencies. Transient faults and software errors were not considered. A different analyst was responsible for each technique. Significantly more time and effort were required to learn performability analysis than the fault tree method. Performability is inherently as accurate as fault tree analysis. For the sample problems, fault trees were more practical and less time consuming to apply, while performability required less ingenuity and was more checkable. Performability offers some advantages for evaluating very complex problems.

Hitt, E. F.↗