Design Space Exploration for MDO on a Teraflop Computer
The objective of this project was to make efficient use of massively parallel computation for exploration of high-dimensional aircraft-configuration design space.
Grossman, Bernard↗
SEARCH · Engineering Papers
Results for “high dimensionality”
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.
Accuracy Assessment of Response Surface Approximations for Supersonic Turbine Design
There is a growing trend to employ CFD tools to supply the necessary information for design optimization of fluid dynamics components/systems. Such results are prone to uncertainties due to reasons including discretization. errors, incomplete convergence of computational procedures, and errors associated with physical models such as turbulence closures. Based on this type of information, gradient-based optimization algorithms often suffer from the noisy calculations, which can seriously compromise the outcome. Similar problems arise from the experimental measurements. Global optimization techniques, such as those based on the response surface (RS) concept are becoming popular in part because they can overcome some of these barriers. However, there are also fundamental issues related to such global optimization technique such as RS. For example, in high dimensional design spaces, typically only a small number of function evaluations are available due to computational and experimental costs. On the other hand, complex features of the design variables do not allow one to model the global characteristics of the design space with simple quadratic polynomials. Consequently a main challenge is to reduce the size of the region where we fit the RS, or make it more accurate in the regions where the optimum is likely to reside. Response Surface techniques using either polynomials or and Neural Network (NN) methods offer designers alternatives to conduct design optimization. The RS technique employs statistical and numerical techniques to establish the relationship between design variables and objective/constraint functions, typically using polynomials. In this study, we aim at addressing issues related to the following questions: (1) How to identify outliers associated with a given RS representation and improve the RS model via appropriate treatments? (2) How to focus on selected design data so that RS can give better performance in regions critical to design optimization? (3) How to combine NN and polynomial techniques for improving the accuracy of the RS model?
Papila, Nilay↗
Fast 3D Net Expeditions: Tools for Effective Scientific Collaboration on the World Wide Web
Two new technologies, the FASTexpedition and Remote FAST, have been developed that provide remote, 3D (three dimensional), high resolution, dynamic, interactive viewing of scientific data. The FASTexpedition permits one to access scientific data from the World Wide Web, take guided expeditions through the data, and continue with self controlled expeditions through the data. Remote FAST permits collaborators at remote sites to simultaneously view an analysis of scientific data being controlled by one of the collaborators. Control can be transferred between sites. These technologies are now being used for remote collaboration in joint university, industry, and NASA projects. Also, NASA Ames Research Center has initiated a project to make scientific data and guided expeditions through the data available as FASTexpeditions on the World Wide Web for educational purposes. Previously, remote visualization of dynamic data was done using video format (transmitting pixel information) such as video conferencing or MPEG (Motion Picture Expert Group) movies on the Internet. The concept for this new technology is to send the raw data (e.g., grids, vectors, and scalars) along with viewing scripts over the Internet and have the pixels generated by a visualization tool running on the viewers local workstation. The visualization tool that is currently used is FAST (Flow Analysis Software Toolkit). The advantages of this new technology over using video format are: (1) The visual is much higher in resolution (1280x1024 pixels with 24 bits of color) than typical video format transmitted over the network. (2) The form of the visualization can be controlled interactively (because the viewer is interactively controlling the visualization tool running on his workstation). (3) A rich variety of guided expeditions through the data can be included easily. (4) A capability is provided for other sites to see a visual analysis of one site as the analysis is interactively performed. Control of the analysis can be passed from site to site. (5) The scenes can be viewed in 3D using stereo vision. (6) The network bandwidth for the visualization using this new technology is much smaller than when using video format. (The measured peak bandwidth used was 1 Kbit/sec whereas the measured bandwidth for a small video picture was 500 Kbits/sec.) This talk will illustrate the use of these new technologies and present a proposal for using these technologies to improve science education.
Watson, Val↗
Hyperspectral Image Classification using a Self-Organizing Map
The use of hyperspectral data to determine the abundance of constituents in a certain portion of the Earth's surface relies on the capability of imaging spectrometers to provide a large amount of information at each pixel of a certain scene. Today, hyperspectral imaging sensors are capable of generating unprecedented volumes of radiometric data. The Airborne Visible/Infrared Imaging Spectrometer (AVIRIS), for example, routinely produces image cubes with 224 spectral bands. This undoubtedly opens a wide range of new possibilities, but the analysis of such a massive amount of information is not an easy task. In fact, most of the existing algorithms devoted to analyzing multispectral images are not applicable in the hyperspectral domain, because of the size and high dimensionality of the images. The application of neural networks to perform unsupervised classification of hyperspectral data has been tested by several authors and also by us in some previous work. We have also focused on analyzing the intrinsic capability of neural networks to parallelize the whole hyperspectral unmixing process. The results shown in this work indicate that neural network models are able to find clusters of closely related hyperspectral signatures, and thus can be used as a powerful tool to achieve the desired classification. The present work discusses the possibility of using a Self Organizing neural network to perform unsupervised classification of hyperspectral images. In sections 3 and 4, the topology of the proposed neural network and the training algorithm are respectively described. Section 5 provides the results we have obtained after applying the proposed methodology to real hyperspectral data, described in section 2. Different parameters in the learning stage have been modified in order to obtain a detailed description of their influence on the final results. Finally, in section 6 we provide the conclusions at which we have arrived.
Martinez, P.↗
Demonstrations at Supercomputing 1998
Contents of the papers include: Visualization techniques- The next generation, Small-machine visualizing very big data sets, Public display of Lunar Prospector data, Steering molecules via Feel, High-dimensional data browser, Multi-source visualization, Interactive surface flow visualization, and the Virtual Windtunnel.
Bryson, Steve↗
Mining Distance Based Outliers in Near Linear Time with Randomization and a Simple Pruning Rule
Defining outliers by their distance to neighboring examples is a popular approach to finding unusual examples in a data set. Recently, much work has been conducted with the goal of finding fast algorithms for this task. We show that a simple nested loop algorithm that in the worst case is quadratic can give near linear time performance when the data is in random order and a simple pruning rule is used. We test our algorithm on real high-dimensional data sets with millions of examples and show that the near linear scaling holds over several orders of magnitude. Our average case analysis suggests that much of the efficiency is because the time to process non-outliers, which are the majority of examples, does not depend on the size of the data set.
Bay, Stephen D.↗
The Hyperwall
This paper presents the hyperwall, a visualization cluster that uses coordinated visualizations for interactive exploration of multidimensional data and simulations. The system strongly leverages the human eye-brain system with a generous 7x7 array offlat panel LCD screens powered by a beowulf clustel: With each screen backed by a workstation class PC, graphic and compute intensive applications can be applied to a broad range of data. Navigational tools are presented that allow for investigation of high dimensional spaces.
Biegel, Bryan A.↗
Multi-Agent Methods for the Configuration of Random Nanocomputers
As computational devices continue to shrink, the cost of manufacturing such devices is expected to grow exponentially. One alternative to the costly, detailed design and assembly of conventional computers is to place the nano-electronic components randomly on a chip. The price for such a trivial assembly process is that the resulting chip would not be programmable by conventional means. In this work, we show that such random nanocomputers can be adaptively programmed using multi-agent methods. This is accomplished through the optimization of an associated high dimensional error function. By representing each of the independent variables as a reinforcement learning agent, we are able to achieve convergence must faster than with other methods, including simulated annealing. Standard combinational logic circuits such as adders and multipliers are implemented in a straightforward manner. In addition, we show that the intrinsic flexibility of these adaptive methods allows the random computers to be reconfigured easily, making them reusable. Recovery from faults is also demonstrated.
Lawson, John W.↗
Gene Expression Dynamics Inspector (GEDI): for integrative analysis of expression profiles
Genome-wide expression profiles contain global patterns that evade visual detection in current gene clustering analysis. Here, a Gene Expression Dynamics Inspector (GEDI) is described that uses self-organizing maps to translate high-dimensional expression profiles of time courses or sample classes into animated, coherent and robust mosaics images. GEDI facilitates identification of interesting patterns of molecular activity simultaneously across gene, time and sample space without prior assumption of any structure in the data, and then permits the user to retrieve genes of interest. Important changes in genome-wide activities may be quickly identified based on 'Gestalt' recognition and hence, GEDI may be especially useful for non-specialist end users, such as physicians. AVAILABILITY: GEDI v1.0 is written in Matlab, and binary Matlab.dll files which require Matlab to run can be downloaded for free by academic institutions at http://www.chip.org/~ge/gedihome.html Supplementary information: http://www.chip.org/~ge/gedihome.html.
Database Management Systems↗
Development and Testing of Data Mining Algorithms for Earth Observation
The new algorithms developed under this project included a principled procedure for classification of objects, events or circumstances according to a target variable when a very large number of potential predictor variables is available but the number of cases that can be used for training a classifier is relatively small. These "high dimensional" problems require finding a minimal set of variables -called the Markov Blanket-- sufficient for predicting the value of the target variable. An algorithm, the Markov Blanket Fan Search, was developed, implemented and tested on both simulated and real data in conjunction with a graphical model classifier, which was also implemented. Another algorithm developed and implemented in TETRAD IV for time series elaborated on work by C. Granger and N. Swanson, which in turn exploited some of our earlier work. The algorithms in question learn a linear time series model from data. Given such a time series, the simultaneous residual covariances, after factoring out time dependencies, may provide information about causal processes that occur more rapidly than the time series representation allow, so called simultaneous or contemporaneous causal processes. Working with A. Monetta, a graduate student from Italy, we produced the correct statistics for estimating the contemporaneous causal structure from time series data using the TETRAD IV suite of algorithms. Two economists, David Bessler and Kevin Hoover, have independently published applications using TETRAD style algorithms to the same purpose. These implementations and algorithmic developments were separately used in two kinds of studies of climate data: Short time series of geographically proximate climate variables predicting agricultural effects in California, and longer duration climate measurements of temperature teleconnections.
Glymour, Clark↗
Discovering System Health Anomalies Using Data Mining Techniques
We present a data mining framework for the analysis and discovery of anomalies in high-dimensional time series of sensor measurements that would be found in an Integrated System Health Monitoring system. We specifically treat the problem of discovering anomalous features in the time series that may be indicative of a system anomaly, or in the case of a manned system, an anomaly due to the human. Identification of these anomalies is crucial to building stable, reusable, and cost-efficient systems. The framework consists of an analysis platform and new algorithms that can scale to thousands of sensor streams to discovers temporal anomalies. We discuss the mathematical framework that underlies the system and also describe in detail how this framework is general enough to encompass both discrete and continuous sensor measurements. We also describe a new set of data mining algorithms based on kernel methods and hidden Markov models that allow for the rapid assimilation, analysis, and discovery of system anomalies. We then describe the performance of the system on a real-world problem in the aircraft domain where we analyze the cockpit data from aircraft as well as data from the aircraft propulsion, control, and guidance systems. These data are discrete and continuous sensor measurements and are dealt with seamlessly in order to discover anomalous flights. We conclude with recommendations that describe the tradeoffs in building an integrated scalable platform for robust anomaly detection in ISHM applications.
Sriastava, Ashok, N.↗
A Comparative Study of Probability Collectives Based Multi-agent Systems and Genetic Algorithms
We compare Genetic Algorithms (GA's) with Probability Collectives (PC), a new framework for distributed optimization and control. In contrast to GA's, PC-based methods do not update populations of solutions. Instead they update an explicitly parameterized probability distribution p over the space of solutions. That updating of p arises as the optimization of a functional of p. The functional is chosen so that any p that optimizes it should be p peaked about good solutions. The PC approach works in both continuous and discrete problems. It does not suffer from the resolution limitation of the finite bit length encoding of parameters into GA alleles. It also has deep connections with both game theory and statistical physics. We review the PC approach using its motivation as the information theoretic formulation of bounded rationality for multi-agent systems. It is then compared with GA's on a diverse set of problems. To handle high dimensional surfaces, in the PC method investigated here p is restricted to a product distribution. Each distribution in that product is controlled by a separate agent. The test functions were selected for their difficulty using either traditional gradient descent or genetic algorithms. On those functions the PC-based approach significantly outperforms traditional GA's in both rate of descent, trapping in false minima, and long term optimization.
Huang, Chien-Feng↗
Visualizing Mercer Kernel feature spaces via kernelized locally-linear embeddings
A new technique for projecting high-dimensional data to low-dimensional spaces, called locally linear embedding (LLE) has recently been introduced.
locally-linear embedding kernel methods↗
On the Application of a Response Surface Technique to Analyze Roll-over Stability of Capsules with Airbags Using LS-Dyna
As NASA moves towards developing technologies needed to implement its new Exploration program, studies conducted for Apollo in the 1960's to understand the rollover stability of capsules landing are being revisited. Although rigid body kinematics analyses of the roll-over behavior of capsules on impact provided critical insight to the Apollo problem, extensive ground test programs were also used. For the new Orion spacecraft being developed to implement today's Exploration program, new air-bag designs have improved sufficiently for NASA to consider their use to mitigate landing loads to ensure crew safety and to enable re-usability of the capsule. Simple kinematics models provide only limited understanding of the behavior of these air bag systems, and more sophisticated tools must be used. In particular, NASA and its contractors are using the LS-Dyna nonlinear simulation code for impact response predictions of the full Orion vehicle with air bags by leveraging the extensive air bag prediction work previously done by the automotive industry. However, even in today's computational environment, these analyses are still high-dimensional, time consuming, and computationally intensive. To alleviate the computational burden, this paper presents an approach that uses deterministic sampling techniques and an adaptive response surface method to not only use existing LS-Dyna solutions but also to interpolate from LS-Dyna solutions to predict the stability boundaries for a capsule on airbags. Results for the stability boundary in terms of impact velocities, capsule attitude, impact plane orientation, and impact surface friction are discussed.
Horta, Lucas G.↗
A Computational Approach for Probabilistic Analysis of Water Impact Simulations
NASA's development of new concepts for the Crew Exploration Vehicle Orion presents many similar challenges to those worked in the sixties during the Apollo program. However, with improved modeling capabilities, new challenges arise. For example, the use of the commercial code LS-DYNA, although widely used and accepted in the technical community, often involves high-dimensional, time consuming, and computationally intensive simulations. The challenge is to capture what is learned from a limited number of LS-DYNA simulations to develop models that allow users to conduct interpolation of solutions at a fraction of the computational time. This paper presents a description of the LS-DYNA model, a brief summary of the response surface techniques, the analysis of variance approach used in the sensitivity studies, equations used to estimate impact parameters, results showing conditions that might cause injuries, and concluding remarks.
Horta, Lucas G.↗
Physical Principle for Generation of Randomness
A physical principle (more precisely, a principle that incorporates mathematical models used in physics) has been conceived as the basis of a method of generating randomness in Monte Carlo simulations. The principle eliminates the need for conventional random-number generators. The Monte Carlo simulation method is among the most powerful computational methods for solving high-dimensional problems in physics, chemistry, economics, and information processing. The Monte Carlo simulation method is especially effective for solving problems in which computational complexity increases exponentially with dimensionality. The main advantage of the Monte Carlo simulation method over other methods is that the demand on computational resources becomes independent of dimensionality. As augmented by the present principle, the Monte Carlo simulation method becomes an even more powerful computational method that is especially useful for solving problems associated with dynamics of fluids, planning, scheduling, and combinatorial optimization. The present principle is based on coupling of dynamical equations with the corresponding Liouville equation. The randomness is generated by non-Lipschitz instability of dynamics triggered and controlled by feedback from the Liouville equation. (In non-Lipschitz dynamics, the derivatives of solutions of the dynamical equations are not required to be bounded.)
Zak, Michail↗
Algorithms for Spectral Decomposition with Applications to Optical Plume Anomaly Detection
The analysis of spectral signals for features that represent physical phenomenon is ubiquitous in the science and engineering communities. There are two main approaches that can be taken to extract relevant features from these high-dimensional data streams. The first set of approaches relies on extracting features using a physics-based paradigm where the underlying physical mechanism that generates the spectra is used to infer the most important features in the data stream. We focus on a complementary methodology that uses a data-driven technique that is informed by the underlying physics but also has the ability to adapt to unmodeled system attributes and dynamics. We discuss the following four algorithms: Spectral Decomposition Algorithm (SDA), Non-Negative Matrix Factorization (NMF), Independent Component Analysis (ICA) and Principal Components Analysis (PCA) and compare their performance on a spectral emulator which we use to generate artificial data with known statistical properties. This spectral emulator mimics the real-world phenomena arising from the plume of the space shuttle main engine and can be used to validate the results that arise from various spectral decomposition algorithms and is very useful for situations where real-world systems have very low probabilities of fault or failure. Our results indicate that methods like SDA and NMF provide a straightforward way of incorporating prior physical knowledge while NMF with a tuning mechanism can give superior performance on some tests. We demonstrate these algorithms to detect potential system-health issues on data from a spectral emulator with tunable health parameters.
Srivastava, Askok N.↗
Multiple Kernel Learning for Heterogeneous Anomaly Detection: Algorithm and Aviation Safety Case Study
The world-wide aviation system is one of the most complex dynamical systems ever developed and is generating data at an extremely rapid rate. Most modern commercial aircraft record several hundred flight parameters including information from the guidance, navigation, and control systems, the avionics and propulsion systems, and the pilot inputs into the aircraft. These parameters may be continuous measurements or binary or categorical measurements recorded in one second intervals for the duration of the flight. Currently, most approaches to aviation safety are reactive, meaning that they are designed to react to an aviation safety incident or accident. In this paper, we discuss a novel approach based on the theory of multiple kernel learning to detect potential safety anomalies in very large data bases of discrete and continuous data from world-wide operations of commercial fleets. We pose a general anomaly detection problem which includes both discrete and continuous data streams, where we assume that the discrete streams have a causal influence on the continuous streams. We also assume that atypical sequence of events in the discrete streams can lead to off-nominal system performance. We discuss the application domain, novel algorithms, and also discuss results on real-world data sets. Our algorithm uncovers operationally significant events in high dimensional data streams in the aviation industry which are not detectable using state of the art methods
Das, Santanu↗