Open Access. Powered by Scholars. Published by Universities.®
Articles 1 - 22 of 22
Full-Text Articles in Entire DC Network
Neural Extensions To Robust Parameter Design, Bernard Jacob Loeffelholz
Neural Extensions To Robust Parameter Design, Bernard Jacob Loeffelholz
Theses and Dissertations
Robust parameter design (RPD) is implemented in systems in which a user wants to minimize the variance of a system response caused by uncontrollable factors while obtaining a consistent and reliable system response over time. We propose the use of artificial neural networks to compensate for highly non-linear problems that quadratic regression fails to accurately model. RPD is conducted under the assumption that the relationship between system response and controllable and uncontrollable variables does not change over time. We propose a methodology to find a new set of settings that will be robust to moderate system degradation while remaining robust …
Evolutionary Artificial Neural Network Weight Tuning To Optimize Decision Making For An Abstract Game, Corey M. Miller
Evolutionary Artificial Neural Network Weight Tuning To Optimize Decision Making For An Abstract Game, Corey M. Miller
Theses and Dissertations
Abstract strategy games present a deterministic perfect information environment with which to test the strategic capabilities of artificial intelligence systems. With no unknowns or random elements, only the competitors’ performances impact the results. This thesis takes one such game, Lines of Action, and attempts to develop a competitive heuristic. Due to the complexity of Lines of Action, artificial neural networks are utilized to model the relative values of board states. An application, pLoGANN (Parallel Lines of Action with Genetic Algorithm and Neural Networks), is developed to train the weights of this neural network by implementing a genetic algorithm over a …
A Comparison Of Main Rotor Smoothing Adjustments Using Linear And Neural Network Algorithms, Nathan A. Miller
A Comparison Of Main Rotor Smoothing Adjustments Using Linear And Neural Network Algorithms, Nathan A. Miller
Theses and Dissertations
Helicopter main rotor smoothing is a maintenance procedure that is routinely performed to minimize airframe vibrations induced by non-uniform mass and/or aerodynamic distributions in the main rotor system. This important task is both time consuming and expensive, so improvements to the process have long been sought. Traditionally, vibrations have been minimized by calculating adjustments based on an assumed linear relationship between adjustments and vibration response. In recent years, artificial neural networks have been trained to recognize non-linear mappings between adjustments and vibration response. This research was conducted in order observe the character of the adjustment mapping of the Vibration Management …
Software Development Outsourcing Decision Support Tool With Neural Network Learning, James D. Newberry
Software Development Outsourcing Decision Support Tool With Neural Network Learning, James D. Newberry
Theses and Dissertations
The Air Force (AF) needs an evolving software tool for guiding decision makers through the complexities of software outsourcing. Previous research identified specific outsourcing strategies and linked them to goals and consequences through a variety of relationship rules. These strategies and relationship rules were inserted into a decision support tool. Since that time, more historical data and outsourcing literature has been collected thus necessitating an update to such a tool. As the number of software outsourcing projects are completed, the AF must capture the outsourcing decision experiences which guided the projects and their outcomes. In order to efficiently incorporate this …
An Investigation Of The Effects Of Correlation, Autocorrelation, And Sample Size In Classifier Fusion, Nathan J. Leap
An Investigation Of The Effects Of Correlation, Autocorrelation, And Sample Size In Classifier Fusion, Nathan J. Leap
Theses and Dissertations
This thesis extends the research found in Storm, Bauer, and Oxley, 2003. Data correlation effects and sample size effects on three classifier fusion techniques and one data fusion technique were investigated. Identification System Operating Characteristic Fusion (Haspert, 2000), the Receiver Operating Characteristic Within Fusion method (Oxley and Bauer, 2002), and a Probabilistic Neural Network were the three classifier fusion techniques; a Generalized Regression Neural Network was the data fusion technique. Correlation was injected into the data set both within a feature set (autocorrelation) and across feature sets for a variety of classification problems, and sample size was varied throughout. Total …
An Integrated Architecture And Feature Selection Algorithm For Radial Basis Neural Networks, Timothy D. Flietstra
An Integrated Architecture And Feature Selection Algorithm For Radial Basis Neural Networks, Timothy D. Flietstra
Theses and Dissertations
There are two basic ways to control an Unmanned Combat Aerial Vehicle (UCAV) as it searches for targets: allow the UCAV to act autonomously or employ man-in-the-loop control. There are also two target sets of interest: fixed or mobile targets. This research focuses on UCAV-based targeting of mobile targets using man-in-the-loop control. In particular, the interest is in how levels of satellite signal latency or signal degradation affect the ability to accurately track, target, and attack mobile targets. This research establishes a weapon effectiveness model assessing targeting inaccuracies as a function of signal latency and/or signal degradation. The research involved …
Predicting Launch Pad Winds At The Kennedy Space Center With A Neural Network Model, Steven J. Storch
Predicting Launch Pad Winds At The Kennedy Space Center With A Neural Network Model, Steven J. Storch
Theses and Dissertations
This thesis uses neural networks to forecast winds at the Kennedy Space Center and the Cape Canaveral Air Station launch pads. Variables are developed from WINDS tower observations, surface and buoy observations, and an upper-air sounding. From these variables, a smaller set of predictive inputs is chosen using a signal-to-noise variable screening method. A neural network is then trained to forecast launch pad winds from the inputs. The network forecasts are compared to persistence, and peak wind predictions are found skillful compared to persistence. An ensemble modeling technique using Toth's and Kalnay's breeding of growing modes method is explored with …
Autonomous Construction Of Multi Layer Perceptron Neural Networks, Thomas F. Rathbun
Autonomous Construction Of Multi Layer Perceptron Neural Networks, Thomas F. Rathbun
Theses and Dissertations
The construction of Multi Layer Perceptron (MLP) neural networks for classification is explored. A novel algorithm is developed, the MLP Iterative Construction Algorithm (MICA), that designs the network architecture as it trains the weights of the hidden layer nodes. The architecture can be optimized on training set classification accuracy, whereby it always achieves 100% classification accuracies, or it can be optimized for generalization. The test results for MICA compare favorably with those of backpropagation on some data sets and far surpasses backpropagation on others while requiring less FLOPS to train. Feature selection is enhanced by MICA because it affords the …
Physiologically-Based Vision Modeling Applications And Gradient Descent-Based Parameter Adaptation Of Pulse Coupled Neural Networks, Randy P. Broussard
Physiologically-Based Vision Modeling Applications And Gradient Descent-Based Parameter Adaptation Of Pulse Coupled Neural Networks, Randy P. Broussard
Theses and Dissertations
In this research, pulse coupled neural networks (PCNNs) are analyzed and evaluated for use in primate vision modeling. An adaptive PCNN is developed that automatically sets near-optimal parameter values to achieve a desired output. For vision modeling, a physiologically motivated vision model is developed from current theoretical and experimental biological data. The biological vision processing principles used in this model, such as spatial frequency filtering, competitive feature selection, multiple processing paths, and state dependent modulation are analyzed and implemented to create a PCNN based feature extraction network. This network extracts luminance, orientation, pitch, wavelength, and motion, and can be cascaded …
Experiments In Aggregating Air Ordnance Effectiveness Data For The Tacwar Model, James E. Parker
Experiments In Aggregating Air Ordnance Effectiveness Data For The Tacwar Model, James E. Parker
Theses and Dissertations
An interactive MS Access&trademark; based application that aggregates the output of the SABSEL model for input into the TACWAR model is developed. The application was developed following efforts to create a functional approximation of the SABSEL data using neural networks, statistical networks, and traditional statistical techniques. These approximations were compared to a look-up table methodology on the basis of accuracy, (RMSE
Pulse Coupled Neural Networks For The Segmentation Of Magnetic Resonance Brain Images, Shane L. Abrahamson
Pulse Coupled Neural Networks For The Segmentation Of Magnetic Resonance Brain Images, Shane L. Abrahamson
Theses and Dissertations
This research develops an automated method for segmenting Magnetic Resonance (MR) brain images based on Pulse Coupled Neural Networks (PCNN). MR brain image segmentation has proven difficult, primarily due to scanning artifacts such as interscan and intrascan intensity inhomogeneities. The method developed and presented here uses a PCNN to both filter and segment MR brain images. The technique begins by preprocessing images with a PCNN filter to reduce scanning artifacts. Images are then contrast enhanced via histogram equalization. Finally, a PCNN is used to segment the images to arrive at the final result. Modifications to the original PCNN model are …
A Neural Network Approach To The Prediction And Confidence Assignation Of Nonlinear Time Series Classifications, Erin S. Heim
A Neural Network Approach To The Prediction And Confidence Assignation Of Nonlinear Time Series Classifications, Erin S. Heim
Theses and Dissertations
This thesis uses multiple layer perceptrons (MLP) neural networks and Kohonen clustering networks to predict and assign confidence to nonlinear time series classifications. The nonlinear time series used for analysis is the Standard and Poor's 100 (S&P 100) index. The target prediction is classification of the daily index change. Financial indicators were evaluated to determine the most useful combination of features for input into the networks. After evaluation it was determined that net changes in the index over time and three short-term indicators result in better accuracy. A back-propagation trained MLP neural network was then trained with these features to …
Temporal Influence On Awareness, Don E. Hill
Temporal Influence On Awareness, Don E. Hill
Theses and Dissertations
Grossberg's Motion Oriented Contrast Filter (MOC) was extensively analyzed (7). The output from the filter's "global motion" neuronal layer was compared to a noncausal post-processing filter developed by AFIT. Both filters were shown to incorporate a weighted, noncausal temporal range of input data in processed output. The global motion framework was then implemented using a physiologically motivated pulsed neural model - the Pulse Coupled Neural Network (PCNN). By incorporating both spatial and temporal data, the PCNN was shown to exhibit a common visual illusion, apparent motion. The existence of a physiological temporal processing range was further investigated through implementation of …
Semantic Interpretation Of An Artificial Neural Network, Stanley D. Kinderknecht
Semantic Interpretation Of An Artificial Neural Network, Stanley D. Kinderknecht
Theses and Dissertations
Recent advances in machine learning theory have opened the door for applications to many difficult problem domains. One area that has achieved great success for stock market analysis/prediction is artificial neural networks. However, knowledge embedded in the neural network is not easily translated into symbolic form. Recent research, exploring the viability of merging artificial neural networks with traditional rule-based expert systems, has achieved limited success. In particular, extracting production (IF.. THEN) rules from a trained neural net based on connection weights provides a valid set of rules only when neuron outputs are close to 0 or 1 (e.g. the output …
The Mathematics Of Measuring Capabilities Of Artificial Neural Networks, Martha A. Carter
The Mathematics Of Measuring Capabilities Of Artificial Neural Networks, Martha A. Carter
Theses and Dissertations
Researchers rely on the mathematics of Vapnik and Chervonenkis to capture quantitatively the capabilities of specific artificial neural network (ANN) architectures. The quantifier is known as the V-C dimension, and is defined on functions or sets. Its value is the largest cardinality 1 of a set of vectors in Rd such that there is at least one set of vectors of cardinality 1 such that all dichotomies of that set into two sets can be implemented by the function or set. Stated another way, the V-C dimension of a set of functions is the largest cardinality of a set, such …
Nonlinear Time Series Analysis, James A. Stewart
Nonlinear Time Series Analysis, James A. Stewart
Theses and Dissertations
This thesis applies neural network feature selection techniques to multivariate time series data to improve prediction of a target time series. Two approaches to feature selection are used. First, a subset enumeration method is used to determine which financial indicators are most useful for aiding in prediction of the S&P 500 futures daily price. The candidate indicators evaluated include RSI, Stochastics and several moving averages. Results indicate that the Stochastics and RSI indicators result in better prediction results than the moving averages. The second approach to feature selection is calculation of individual saliency metrics. A new decision boundary-based individual saliency …
Embedology And Neural Estimation For Time Series Prediction, Robert E. Garza
Embedology And Neural Estimation For Time Series Prediction, Robert E. Garza
Theses and Dissertations
Time series prediction has widespread application, ranging from predicting the stock market to trying to predict future locations of scud missiles. Recent work by Sauer and Casdagli has developed into the embedology theorem, which sets forth the procedures for state space manipulation and reconstruction for time series prediction. This includes embedding the time series into a higher dimensional space in order to form an attractor, a structure defined by the embedded vectors. Embedology is combined with neural technologies in an effort to create a more accurate prediction algorithm. These algorithms consist of embedology, neural networks, Euclidean space nearest neighbors, and …
Feature And Model Selection In Feedforward Neural Networks, Jean M. Steppe
Feature And Model Selection In Feedforward Neural Networks, Jean M. Steppe
Theses and Dissertations
This research advances feature and model selection for feedforward neural networks. Feature selection involves determining a good feature subset given a set of candidate features. Model selection involves determining an appropriate architecture number of middle nodes for the neural network. Specific advances are made in neural network feature saliency metrics used for evaluating or ranking features, statistical identification of irrelevant noisy features, and statistical investigation of reduced neural network architectures and reduced feature subsets. New feature saliency metrics are presented which provide a more succinct quantitative measure of a features importance than other similar metrics. A catalogue of feature saliency …
Subgrouped Real Time Recurrent Learning Neural Networks, Jeffrey S. Dean
Subgrouped Real Time Recurrent Learning Neural Networks, Jeffrey S. Dean
Theses and Dissertations
A subgrouped Real Time Recurrent Learning (RTRL) network was evaluated. The one layer net successfully learns the XOR problem, and can be trained to perform time dependent functions. The net was tested as a predictor on the behavior of a signal, based on past behavior. While the net was not able to predict the signal's future behavior, it tracked the signal closely. The net was also tested as a classifier for time varying phenomena; for the differentiation of five classes of vehicle images based on features extracted from the visual information. The net achieved a 99.2% accuracy in recognizing the …
A Fortran Based Learning System Using Multilayer Back-Propagation Neural Network Techniques, Gregory L. Reinhart
A Fortran Based Learning System Using Multilayer Back-Propagation Neural Network Techniques, Gregory L. Reinhart
Theses and Dissertations
An interactive computer system which allows the researcher to build an optimal neural network structure quickly, is developed and validated. This system assumes a single hidden layer perceptron structure and uses the back- propagation training technique. The software enables the researcher to quickly define a neural network structure, train the neural network, interrupt training at any point to analyze the status of the current network, re-start training at the interrupted point if desired, and analyze the final network using two- dimensional graphs, three-dimensional graphs, confusion matrices and saliency metrics. A technique for training, testing, and validating various network structures and …
Predicting Nonlinear Time Series, James C. Gainey Jr.
Predicting Nonlinear Time Series, James C. Gainey Jr.
Theses and Dissertations
Predicting future values of a time series has many practical uses in real-time signal processing and understanding. This thesis implements an Adaptive Time Delay Neural Network ATNN capable of user-defined degeneration to the more common Time Delay Neural Network TDNN. Time delays along axons or at the synapses, which vary in biological systems, motivate this research. The ATNNTDNN test results and time series prediction capabilities are compared to those of the Real-Time Recurrent Learning RTRL algorithm. To show the advantages and disadvantages of using TDNN and ATNN for prediction versus the RTRL, the networks were applied to two problems incommensurate …
Color Image Segmentation, Kimberley A. Mccrae
Color Image Segmentation, Kimberley A. Mccrae
Theses and Dissertations
The most difficult stage of automated target recognition ATR is segmentation. Current AFIT segmentation problems include faces and tactical targets previous efforts to segment these objects have used intensity and motion cues. This thesis develops a color preprocessing scheme to be used with the other segmentation techniques. A neural network is trained to identify the color of a desired object, eliminating all but that color from the scene. Gabor correlations and 2D wavelet transformations will be performed on stationary images and 3D wavelet transforms on multispectral data will incorporate color and motion detection into the machine visual system. The thesis …