End-to-end machine learning for experimental physics: using simulated data to train a neural network for object detection in video microscopy
Explore the source record for details and available documents.
Eric N. Minor↗
SEARCH · Search NASA
Results for “experimental 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.
Automatic cataloguing and characterization of Earth science data using SE-trees
In the future, NASA's Earth Observing System (EOS) platforms will produce enormous amounts of remote sensing image data that will be stored in the EOS Data Information System. For the past several years, the Intelligent Data Management group at Goddard's Information Science and Technology Office has been researching techniques for automatically cataloguing and characterizing image data (ADCC) from EOS into a distributed database. At the core of the approach, scientists will be able to retrieve data based upon the contents of the imagery. The ability to automatically classify imagery is key to the success of contents-based search. We report results from experiments applying a novel machine learning framework, based on Set-Enumeration (SE) trees, to the ADCC domain. We experiment with two images: one taken from the Blackhills region in South Dakota; and the other from the Washington DC area. In a classical machine learning experimentation approach, an image's pixels are randomly partitioned into training (i.e. including ground truth or survey data) and testing sets. The prediction model is built using the pixels in the training set, and its performance is estimated using the testing set. With the first Blackhills image, we perform various experiments achieving an accuracy level of 83.2 percent, compared to 72.7 percent using a Back Propagation Neural Network (BPNN) and 65.3 percent using a Gaussain Maximum Likelihood Classifier (GMLC). However, with the Washington DC image, we were only able to achieve 71.4 percent, compared with 67.7 percent reported for the BPNN model and 62.3 percent for the GMLC.
Rymon, Ron↗
Data Driven Optimization of BN Polymer Composites Using Automated Experimentation and Machine Learning
Explore the source record for details and available documents.
machine learning↗
X-31 Mishap: Lessons Learned
The experimental X-31 High Angle of Attack Research Aircraft crashed during a 1995 test mission flight conducted by NASA at Edwards Air Force Base, California. The pilot lost control of the airplane and was forced to eject, sustaining a permanent back injury that ended his flying career. Prior to this incident the airplane had a perfect record of several hundred non-eventful flights supported by an experienced team. During the subsequent investigation by a mishap committee it was discovered that a series of cascading events contributed to this accident. Some of the identified contributing factors that resulted in this mishap are common to aircraft design and to flight-test in general. The mistakes and the solutions are presented here so that the flight-test community may consider and learn from them. The primary cause of the crash was icing and, ultimately, a complete blockage of the pitot-static nose probe. The icing was caused by a freak weather phenomenon that was neither expected nor known to exist on the day of the mishap. The normal probe had been replaced with a special Kiel probe to allow total pressure measurements of up to 70 degrees angle of attack for flight-test purposes. The Kiel probe did not include a heater, because it was assumed that the airplane would not be flown in the clouds or in conditions conducive to icing. This assumption was later proven to be incorrect. The iced Kiel probe caused incorrect gain scheduling in the flight control system, resulting in an unstable aircraft. This failure was essentially undetected because of a faulty design in the flight control system architecture. There were, however, also a number of other issues that lead up to this situation that never should have happened. This presentation discusses what the issues were that contributed to the incident. After the incident was investigated, some of these issues were addressed and some changes were made. The second X-31 aircraft flew the remainder of the flight tests, and the program was successfully completed without incident. This presentation also shows a video of the mishap including lessons learned, and the changes that were made to resume the flight-test program are presented.
Larson, Richard R.↗
Solving and Learning Soft Temporal Constraints: Experimental Scenario and Examples
Soft temporal constraint problems allow to describe in a natural way scenarios where events happen over time and preferences are associated to event distances and durations. However, sometimes such local preferences are difficult to set, and it may be easier instead to associate preferences to some complete solutions of the problem. To model everything in a uniform way via local preferences only, and also to take advantage of the existing constraint solvers which exploit only local preference use machine learning techniques which learn the local preferences from the global ones. In this paper we describe the existing framework for both solving and learning preferences in temporal constraint problems, the implemented modules, the experimental scenario, and preliminary results on some examples.
Rossi, F.↗
Solving and Learning Soft Temporal Constraints: Experimental Setting and Results
Soft temporal constraints problems allow to describe in a natural way scenarios where events happen over time and preferences are associated to event distances and durations. However, sometimes such local preferences are difficult to set, and it may be easier instead to associate preferences to some complete solutions of the problem. Machine learning techniques can be useful in this respect. In this paper we describe two solvers (one more general and the other one more efficient) for tractable subclasses of soft temporal problems, and we show some experimental results. The random generator used to build the problems on which tests are performed is also described. We also compare the two solvers highlighting the tradeoff between performance and representational power. Finally, we present a learning module and we show its behavior on randomly-generated examples.
Rossi, F.↗
Learning dominance relations in combinatorial search problems
Dominance relations commonly are used to prune unnecessary nodes in search graphs, but they are problem-dependent and cannot be derived by a general procedure. The authors identify machine learning of dominance relations and the applicable learning mechanisms. A study of learning dominance relations using learning by experimentation is described. This system has been able to learn dominance relations for the 0/1-knapsack problem, an inventory problem, the reliability-by-replication problem, the two-machine flow shop problem, a number of single-machine scheduling problems, and a two-machine scheduling problem. It is considered that the same methodology can be extended to learn dominance relations in general.
Yu, Chee-Fen↗
Lana chimpanzee learns to count by 'numath' - A summary of a videotaped experimental report
Computerized training programs whereby an adult female chimpanzee learned to use a joystick to remove from a screen the number of boxes appropriate to the value of a randomly selected Arabic numeral 1, 2, or 3, are studied. Initial training provided a variety of cues, both numeric and otherwise, to support correct performance. In the final test, the ape was correct on over 80 percent of trials in which there was no residual feedback of intratrial events and where only her memory of those events could provide the cue to indicate that she had removed boxes in accordance with the value of the target numbers and should terminate the trial.
Rumbaugh, Duane M.↗
Perceptual learning through optimization of attentional weighting: human versus optimal Bayesian learner
Human performance in visual detection, discrimination, identification, and search tasks typically improves with practice. Psychophysical studies suggest that perceptual learning is mediated by an enhancement in the coding of the signal, and physiological studies suggest that it might be related to the plasticity in the weighting or selection of sensory units coding task relevant information (learning through attention optimization). We propose an experimental paradigm (optimal perceptual learning paradigm) to systematically study the dynamics of perceptual learning in humans by allowing comparisons to that of an optimal Bayesian algorithm and a number of suboptimal learning models. We measured improvement in human localization (eight-alternative forced-choice with feedback) performance of a target randomly sampled from four elongated Gaussian targets with different orientations and polarities and kept as a target for a block of four trials. The results suggest that the human perceptual learning can occur within a lapse of four trials (<1 min) but that human learning is slower and incomplete with respect to the optimal algorithm (23.3% reduction in human efficiency from the 1st-to-4th learning trials). The greatest improvement in human performance, occurring from the 1st-to-2nd learning trial, was also present in the optimal observer, and, thus reflects a property inherent to the visual task and not a property particular to the human perceptual learning mechanism. One notable source of human inefficiency is that, unlike the ideal observer, human learning relies more heavily on previous decisions than on the provided feedback, resulting in no human learning on trials following a previous incorrect localization decision. Finally, the proposed theory and paradigm provide a flexible framework for future studies to evaluate the optimality of human learning of other visual cues and/or sensory modalities.
Non-NASA Center↗
Effects of sleep loss, time of day, and extended mental work on implicit and explicit learning of sequences
Tacit knowledge is part of many professional skills and can be studied experimentally with implicit-learning paradigms. The authors explored the effects of 2 different stressors, loss of sleep and mental fatigue, on implicit learning in a serial-response time (RT) task. In the 1st experiment, 1 night of sleep deprivation was shown to impair implicit but not explicit sequence learning. In the 2nd experiment, no impairment of both types of sequence learning was found after 1.5 hr of mental work. Serial-RT performance, in contrast, suffered from both stressors. These findings suggest that sleep deprivation induces specific risks for automatic, skill-based behavior that are not present in consciously controlled performance.
Learning↗
Modeling to predict pilot performance during CDTI-based in-trail following experiments
A mathematical model was developed of the flight system with the pilot using a cockpit display of traffic information (CDTI) to establish and maintain in-trail spacing behind a lead aircraft during approach. Both in-trail and vertical dynamics were included. The nominal spacing was based on one of three criteria (Constant Time Predictor; Constant Time Delay; or Acceleration Cue). This model was used to simulate digitally the dynamics of a string of multiple following aircraft, including response to initial position errors. The simulation was used to predict the outcome of a series of in-trail following experiments, including pilot performance in maintaining correct longitudinal spacing and vertical position. The experiments were run in the NASA Ames Research Center multi-cab cockpit simulator facility. The experimental results were then used to evaluate the model and its prediction accuracy. Model parameters were adjusted, so that modeled performance matched experimental results. Lessons learned in this modeling and prediction study are summarized.
Sorensen, J. A.↗
ZBLAN Microgravity Study
One of the greatest obstacles with the fluorozirconate ZBLAN (ZrF4-BaF2-LaF3-AIF3-NaF) is the problem of devitrification. Fluoride glasses have a narrow working range and the viscosity is a strong function of temperature. Rates of nucleation and growth of crystals in the glass depend on the viscosity, making these glasses unstable and prone to crystallization. The viscosity of ZBLAN at the drawing temperature is low, usually between two to five poise, so it is difficult to obtain fibers from their preform melts without crystallization. The preforms usually contain heterogeneous nuclei which grow into microcrystallites above the glass transition temperature, T(g). Since microcrystallites in an optical fiber cause extrinsic light scattering losses of the optical signal, fiber drawing must be completed in a short time to minimize the generation of light scattering centers. To keep these losses to a minimum and to fabricate low scattering loss fibers and other optical components, this research deals with the possibility of minimizing crystallite formation by removing the gravitational influence of solutal segregation of the ZBLAN elements. This report reviews the early work on the KC-135 aircraft, the development of the ZBLAN Rocket Experiment, preparations at the White Sands Missile Range, analysis of the flight and ground test results, lessons learned and future experimentation.
Workman, Gary L.↗
A Parametric Geometry Computational Fluid Dynamics (CFD) Study Utilizing Design of Experiments (DOE)
Design of Experiments (DOE) was applied to the LAS geometric parameter study to efficiently identify and rank primary contributors to integrated drag over the vehicles ascent trajectory in an order of magnitude fewer CFD configurations thereby reducing computational resources and solution time. SME s were able to gain a better understanding on the underlying flowphysics of different geometric parameter configurations through the identification of interaction effects. An interaction effect, which describes how the effect of one factor changes with respect to the levels of other factors, is often the key to product optimization. A DOE approach emphasizes a sequential approach to learning through successive experimentation to continuously build on previous knowledge. These studies represent a starting point for expanded experimental activities that will eventually cover the entire design space of the vehicle and flight trajectory.
Rhew, Ray D.↗
Rhesus monkey (Macaca mulatta) complex learning skills reassessed
An automated computerized testing facility is employed to study basic learning and transfer in rhesus monkeys including discrimination learning set and mediational learning. The data show higher performance levels than those predicted from other tests that involved compromised learning with analogous conditions. Advanced transfer-index ratios and positive transfer of learning are identified, and indications of mediational learning strategies are noted. It is suggested that these data are evidence of the effectiveness of the present experimental apparatus for enhancing learning in nonhuman primates.
Washburn, David A.↗
NASA Lessons Learned on Reusable and Expendable Launch Vehicle Operations and Their Application Towards DARPA's Experimental Spaceplane (XS-1) Program
Outline: Selected Historic Shuttle Operations Data; Shuttle Lessons Learned Recommendations for Lower Cost, Operationally Efficient Launch Vehicle Systems; Selected Expendable launch vehicle experiences; Past NASA Launch Vehicle Development Programs, Studies (1985 to present); Discussion: Suggested applications of NASA Lessons Learned to already-baselined contractor XS-1 Phase I concepts.
Reusable launch vehicle↗
Data-Driven Study of Shape Memory Behavior of Multi-Component Ni-Ti Alloys
Ni-Ti based shape memory alloys (SMAs) have found wide-spread use in aerospace, automotive, biomedical, and commercial applications owing to their favorable properties and ease of operation. Especially important for many NASA applications is the ability to tune the martensitic transformation temperature of Ni-Ti alloys by varying the alloy composition and processing conditions. Recently, researchers at NASA have compiled an extensive database of shape memory properties of materials, including over 8,000 multi-component Ni-Ti alloys containing 37 different alloying elements. Using this dataset, machine learning models are trained to predict transformation temperatures, hysteresis, and transformation strain with extremely low mean absolute errors. These models are used to learn relationships between shape memory behavior and input parameters in the composition and processing space. ML predictions are validated through new experiments. The combination of an extensive experimental dataset and accurate learning models, together, make our approach highly suitable for the rapid discovery and design of novel SMAs with targeted properties. We are not aware of any current approaches capable of predicting SMA transformation behavior over such a wide range of compositions and processing conditions.
Shape memory alloys↗
Everybody Needs Somebody Sometimes: Validation of Adaptive Recovery in Robotic Space Operations
This work assesses an adaptive approach to fault recovery in autonomous robotic space operations, which uses indicators of opportunity, such as physiological state measurements and observations of past human assistant performance, to inform future selections. We validated our reinforcement learning approach using data we collected from humans executing simulated mission scenarios. We present a method of structuring human-factors experiments that permits collection of relevant indicator of opportunity and assigned assistance task performance data, as well as evaluation of our adaptive approach, without requiring large numbers of test subjects. Application of our reinforcement learning algorithm to our experimental data shows that our adaptive assistant selection approach can achieve lower cumulative regret compared to existing non-adaptive baseline approaches when using real human data. Our work has applications beyond space robotics to any application where autonomy failures may occur that require external intervention.
Human-centered robotics↗
ACES program - Lessons learned
The ACES Program involved the experimental evaluation of three LSS (large space structures) control design techniques at the LSS GTF (ground test facility) at NASA/MSFC. The three techniques were developed under the ACOSS (active control of space structures) Program specifically for application to LSS. The techniques included FAMESS (filter accommodated model error sensitivity suppression), HAC/LAC (high authority control/low authority control), and positivity. Some of the lessons that have been learned during the course of the ACES program are examined.
Jones, Victoria L.↗