Search Results (1 - 25 of 113 Results)

Sort By  
Sort Dir
 
Results per page  

Kadiyala, AkhilDevelopment and Evaluation of an Integrated Approach to Study In-Bus Exposure Using Data Mining and Artificial Intelligence Methods
Doctor of Philosophy in Engineering, University of Toledo, 2012, Civil Engineering

The objective of this research was to develop and evaluate an integrated approach to model the occupant exposure to in-bus contaminants using the advanced methods of data mining and artificial intelligence. The research objective was accomplished by executing the following steps. Firstly, an experimental field program was implemented to develop a comprehensive one-year database of the hourly averaged in-bus air contaminants (carbon dioxide (CO2), carbon monoxide (CO), nitric oxide (NO), nitrogen dioxide (NO2), sulfur dioxide (SO2), 0.3-0.4 micrometer (¿¿¿¿m) sized particle numbers, 0.4-0.5 ¿¿¿¿m sized particle numbers, particulate matter (PM) concentrations less than 1.0 ¿¿¿¿m (PM1.0), PM concentrations less than 2.5 ¿¿¿¿m (PM2.5), and PM concentrations less than 10.0 ¿¿¿¿m (PM10.0)) and the independent variables (meteorological variables, time-related variables, indoor sources, on-road variables, ventilation settings, and ambient concentrations) that can affect indoor air quality (IAQ). Secondly, a novel approach to characterize in-bus air quality was developed with data mining techniques that incorporated the use of regression trees and the analysis of variance. Thirdly, a new approach to modeling in-bus air quality was established with the development of hybrid genetic algorithm based neural networks (or evolutionary neural networks) with input variables optimized from using the data mining techniques, referred to as the GART approach. Next, the prediction results from the GART approach were evaluated using a comprehensive set of newly developed IAQ operational performance measures. Finally, the occupant exposure to in-bus contaminants was determined by computing the time weighted average (TWA) and comparing them with the recommended IAQ guidelines.

In-bus PM concentrations and sub-micron particle numbers were predominantly influenced by the month/season of the year. In-bus SO2 concentrations were mainly affected by indoor relative humidity (RH) and the month of the year. NO concentrations inside the bus cabin were largely influenced by the indoor RH, while NO2 concentrations primarily varied with the month of the year. Passenger ridership and the month of the year mainly affected the in-bus CO2 concentrations; while the month and sky conditions had a significant impact on CO concentrations within the bus compartment.

The hybrid GART models captured majority of the variance in in-bus contaminant concentrations and performed much better than the traditional artificial neural networks methods of back propagation and radial basis function networks.

Exposure results indicated the average 8-hr. exposure of biodiesel bus occupants to CO2, CO, NO, SO2, and PM2.5 to be 559.67 ppm (¿¿¿¿ 45.01), 18.33 ppm (¿¿¿¿ 9.23), 5.23 ppm (¿¿¿¿ 4.49), 0.13 ppm (¿¿¿¿ 0.01), and 13.75 ¿¿¿¿g/m3 (¿¿¿¿ 4.24), respectively. The statistical significance of the difference in exposure levels to in-bus contaminants were compared during morning, afternoon, and evening/night time periods. There was statistically significant difference only between the morning (driver 1) and the evening/night (driver 3) exposure levels for CO2 and PM2.5. CO levels exceeded the TWA in some months.

Committee:

Dr. Ashok Kumar, PhD (Committee Chair); Dr. Devinder Kaur, PhD (Committee Member); Dr. Cyndee Gruden, PhD (Committee Member); Dr. Defne Apul, PhD (Committee Member); Dr. Farhang Akbar, PhD (Committee Member)

Subjects:

Civil Engineering; Environmental Engineering; Environmental Health

Keywords:

Indoor Air Quality; Public Transportation Buses; Biodiesel; Data Mining; Sensitivity of the Regression Trees; Artificial Neural Networks; Genetic Algorithm Neural Networks; Evolutionary Neural Networks; In-Bus Exposure; Air Quality Model Validation

Gummadi, JayaramA Comparison of Various Interpolation Techniques for Modeling and Estimation of Radon Concentrations in Ohio
Master of Science in Engineering, University of Toledo, 2013, Engineering (Computer Science)
Radon-222 and its parent Radium-226 are naturally occurring radioactive decay products of Uranium-238. The US Environmental Protection Agency (USEPA) attributes about 10 percent of lung cancer cases that is `around 21,000 deaths per year’ in the United States, caused due to indoor radon. The USEPA has categorized Ohio as a Zone 1 state (i.e. the average indoor radon screening level greater than 4 picocuries per liter). In order to implement preventive measures, it is necessary to know radon concentration levels in all the zip codes of a geographic area. However, it is not possible to survey all the zip codes, owing to reasons such as inapproachability. In such places where radon data are unavailable, several interpolation techniques are used to estimate the radon concentrations. This thesis presents a comparison between recently developed interpolation techniques to new techniques such as Support Vector Regression (SVR), and Random Forest Regression (RFR). Recently developed interpolation techniques include Artificial Neural Network (ANN), Knowledge Based Neural Networks (KBNN), Correction-Based Artificial Neural Networks (CBNN) and the conventional interpolation techniques such as Kriging, Local Polynomial Interpolation (LPI), Global Polynomial Interpolation (GPI) and Radial Basis Function (RBF) using the K-fold cross validation method.

Committee:

William Acosta (Committee Chair); Vijay Devabhaktuni (Committee Co-Chair); Ashok Kumar (Committee Member); Rob Green (Committee Member)

Subjects:

Computer Science

Keywords:

artificial neural networks; cross-validation; correction based artificial neural networks; prior knowledge input; source difference; space-mapped neural networks; support vector regression; radon; random forest regression

Ghosh Dastidar, SamanwoyModels of EEG data mining and classification in temporal lobe epilepsy: wavelet-chaos-neural network methodology and spiking neural networks
Doctor of Philosophy, The Ohio State University, 2007, Biomedical Engineering
A multi-paradigm approach integrating three novel computational paradigms: wavelet transforms, chaos theory, and artificial neural networks is developed for EEG-based epilepsy diagnosis and seizure detection. This research challenges the assumption that the EEG represents the dynamics of the entire brain as a unified system. It is postulated that the sub-bands yield more accurate information about constituent neuronal activities underlying the EEG. Consequently, certain changes in EEGs not evident in the original full-spectrum EEG may be amplified when each sub-band is analyzed separately. A novel wavelet-chaos methodology is presented for analysis of EEGs and delta, theta, alpha, beta, and gamma sub-bands of EEGs for detection of seizure and epilepsy. The methodology is applied to three different groups of EEGs: healthy subjects, epileptic subjects during a seizure-free interval (interictal), and epileptic subjects during a seizure (ictal). Two potential markers of abnormality quantifying the non-linear chaotic EEG dynamics are discovered: the correlation dimension and largest Lyapunov exponent. A novel wavelet-chaos-neural network methodology is developed for EEG classification. Along with the aforementioned two parameters, the standard deviation (quantifying the signal variance) is employed for EEG representation. It was discovered that a particular mixed-band feature space consisting of nine parameters and LMBPNN result in the highest classification accuracy (96.7%). To increase the robustness of classification, a novel principal component analysis-enhanced cosine radial basis function neural network classifier is developed. The rearrangement of the input space along the principal components of the data improves the classification accuracy of the cosine radial basis function neural network employed in the second stage significantly. The new classifier is as accurate as LMBPNN and is twice as robust. Next, biologically realistic artificial neural networks are developed to reach the next milestone in artificial intelligence. First, an efficient spiking neural network (SNN) model is presented using three training algorithms: SpikeProp, QuickProp, and RProp. Three measures of performance are investigated: number of convergence epochs, computational efficiency, and classification accuracy. Next, a new Multi-Spiking Neural Network (MuSpiNN) and supervised learning algorithm (Multi-SpikeProp) are developed. Finally, the models are applied to the epilepsy and seizure detection problems to achieve high classification accuracies.

Committee:

Hojjat Adeli (Advisor)

Keywords:

Temporal Lobe Epilepsy; Electroencephalogram (EEG); EEG Classification; Epilepsy Diagnosis; Seizure Detection; Wavelet Transform; Chaos Theory; Artificial Neural Networks; Spiking Neural Networks; Principal Component Analysis; Cosine Radial Basis Function

Pech, Thomas JoelA Deep-Learning Approach to Evaluating the Navigability of Off-Road Terrain from 3-D Imaging
Master of Sciences (Engineering), Case Western Reserve University, 2017, EECS - Computer and Information Sciences
This work investigates a strategy for evaluating the navigability of terrain from 3-D imaging. Labeled training data was automatically generated by running a simulation of a mobile robot nai¨vely exploring a virtual world. During this exploration, sections of terrain were perceived through simulated depth imaging and saved with labels of safe or unsafe, depending on the outcome of the robot's experience driving through the perceived regions. This labeled data was used to train a deep convolutional neural network. Once trained, the network was able to evaluate the safety of perceived regions. The trained network was shown to be effective in achieving safe, autonomous driving through novel, challenging, unmapped terrain.

Committee:

Wyatt Newman (Advisor); Cenk Cavusoglu (Committee Member); Michael Lewicki (Committee Member)

Subjects:

Computer Science; Robotics; Robots

Keywords:

Mobile robots, Autonomous Navigation, Machine Learning, Artificial Neural Networks, Terrain, Simulation, Training Data, Data Generation, Labeling, Classifiers, Convolutional Neural Networks, Point Clouds, Perception, Prediction, Artificial Intelligence

Doboli, SimonaLatent Attractors: A Mechanism for Context-Dependent Information Processing in Biological and Artificial Neural Systems
PhD, University of Cincinnati, 2001, Engineering : Electrical Engineering
The hippocampus is an important area in the brain involved mainly in memory processes. In humans, the hippocampus is essential for the formation and consolidation of memory, and is a primary target of Alzheimer's disease. In other animals, the hippocampus is especially involved in spatial tasks (e.g. navigation). It has been the subject of extensive experimental and theoretical investigation due to its major role in memory and cognition. This thesis focuses mainly on the mechanisms of context-dependent, non-linear spatial information processing in the rodent hippocampus. There is strong experimental evidence that the hippocampus creates and stores cognitive maps of an animal's environment. These maps facilitate path planning and goal-directed behavior – all tasks of great interest in robots as well as animals. However, the mechanisms of spatial information processing in the hippocampus are not completely understood. One aspect of cognitive maps that is not yet clarified is their dependence on the past experience, or context. The first part of the thesis focuses on developing computational models for context-dependent cognitive maps. These models are based on the idea of latent attractors – patterns embedded in recurrent neural networks that influence network dynamics and the response to external inputs without becoming fully manifested themselves. Context-dependent information processing is important not only in animal cognition, but also for problems such as robot navigation, sequence disambiguation, sequential recognition, etc. In the second part of the thesis the biologically inspired concept of latent attractor networks is studied as a general computational paradigm for solving context-dependent problems with neural networks. To gain better understanding of the capabilities of latent attractor networks, a theoretical analysis of their capacity and dynamics is performed. In addition to the model for context-dependence, more comprehensive computational models of the hippocampus are developed to explain specific experimental results such as the effect of changes in the environment on cognitive maps.

Committee:

Ali Minai (Advisor)

Keywords:

context-dependent information processing; attractor neural networks; spatial processing in the hippocampus; analysis of attractor neural networks

Lakumarapu, Shravan KumarCommittee Neural Networks for Image Based Facial Expression Classification System: Parameter Optimization
Master of Science in Engineering, University of Akron, 2010, Biomedical Engineering
There has been a significant growth in the application of Artificial Neural Networks (ANNs) in the medical field including clinical decision support systems which call for high reliability and performance often involving multi-classification problems. Committee Neural Networks (CNNs) were developed for increased reliability and performance and were successfully applied to several multi-classification problems. However, the effect of the number of input parameters on the performance of the CNNs has not been investigated. The purpose of the present study was to investigate this effect on the CNN performance in a multi-classification problem of facial image based mood detection. Kulkarni et al (2009) used 15 parameters to develop a CNN system for classifying a given facial image into one of the six basic moods. Six subsets of parameters, for each image, were extracted from the parameter database used by Kulkarni et al, with an increasing number of parameters (from five to ten). The entire data was divided into training data, initial testing data, and final evaluation data. The training data from the six subsets were used to train six groups of neural networks and these networks were subjected to initial testing using the corresponding initial testing data. CNNs of different sizes were formed for each group by selecting the best performing networks based on the initial testing results. These CNNs were further tested using the initial testing data and the best performing CNN from each group was selected for further evaluation. All the selected committees were further evaluated using the final testing data from subjects not used in training or initial testing. Two approaches were used for converting neural network output into binary: “winner takes all” approach and the “thresholding” approach. The results from both the approaches show that with an increasing number of input parameters, the accuracy increased initially and then decreased. The committee decision was more accurate than individual member networks‟ decision. The highest accuracies obtained were from the CNNs having eight input parameters: 87.9% using the “winner takes all approach” and 87.2% using the “thresholding” approach. In contrast, Kulkarni et al developed a dual committee system consisting of the primary committee and a specialized committee using 15 parameters and reported an accuracy of 90.4%. In the present study, a single committee using only eight input parameters produced similar accuracy. To increase the reliability of neural network based intelligent systems for medical applications, the number of parameters should be optimized.

Committee:

Dr. Narender Reddy, Dr. (Advisor)

Subjects:

Engineering

Keywords:

Artificial Neural Networks; Committee Neural Networks, Artificial Intelligence, decision support systems, optimization

Kulkarni, Saket SFACIAL IMAGE BASED MOOD RECOGNITION USING COMMITTEE NEURAL NETWORKS
Master of Science in Engineering, University of Akron, 2006, Biomedical Engineering
Facial expressions are important in facilitating human communication and interactions. They are also used as an important tool in behavioral studies and medical rehabilitation. Facial image based mood detection techniques may provide a fast and practical approach for noninvasive mood detection. Most of the previous studies of facial expressions recognition have been based on Facial Action Coding System (FACS) developed by Ekman and Freisen in 1978 and identifying different facial muscular actions. Previous neural network based approaches for classification of facial expressions either have used smaller data bases or used the same data for training and testing. The purpose of the present study was to develop an intelligent system for facial image based expression classification using large data bases. Several facial parameters were extracted from facial image and used to train several generalized and specialized neural networks. The best performing generalized and specialized neural networks were recruited into decision making committees forming an integrated committee neural network system. The integrated committee neural network system was evaluated using data obtained from subjects not used in training or initial testing. The system correctly identified the facial expression in 90.426% of the cases, and represents a significant step forward in correctly identifying the facial expression in large facial expressions database.

Committee:

Narender Reddy (Advisor)

Subjects:

Engineering, Biomedical

Keywords:

FACIAL; NEURAL NETWORKS; COMMITTEE NEURAL NETWORKS; Facial expressions; disgust

Horvitz, Richard P.Symbol Grounding Using Neural Networks
PhD, University of Cincinnati, 2012, Engineering and Applied Science: Computer Science and Engineering
The classical approach to artificial intelligence (i.e. symbol manipulation) and the connectionist approach (artificial neural networks) have been criticized for their inadequacies. The philosopher John Searle's Chinese room thought experiment argued that symbolic systems have no understanding of the meaning contained in their representations. The philosophers Jerry Fodor and Zenon Pylyshyn argued that artificial neural networks could not exhibit certain features of human cognition, such as systematicity and composition of representations. We take the view that both of these problems can be solved by a suitable integration of connectionist and symbolic systems. In this work we investigate methods of using artificial neural networks to produce descriptions in propositional and predicate logic. Artificial neural networks are stuctured such that, upon training, simple features of the network correspond directly to either propositional variables in one case, and objects and predicates in the other. In both cases, the methods were tested on character recognition tasks.

Committee:

Raj Bhatnagar, PhD (Committee Chair); Yizong Cheng, PhD (Committee Member); Carla Purdy, PhD (Committee Member); George Purdy, PhD (Committee Member); John Schlipf, PhD (Committee Member)

Subjects:

Computer Science

Keywords:

symbol grounding;artificial neural networks;neural networks;;;;

Shao, YuanlongLearning Sparse Recurrent Neural Networks in Language Modeling
Master of Science, The Ohio State University, 2014, Computer Science and Engineering
In the context of statistical language modeling, we explored the task of learning an Elman network with sparse weight matrices, as a pilot study towards learning a sparsely con-nected fully recurrent neural network, which would be potentially useful in many cases. We also explored how efficient and scalable it can be in practice. In particular, we explored these tasks: (1) We adapted the Iterative Hard Thresholding (IHT) algorithm into the BackPropagation Through Time (BPTT) learning. (2) To accel-erate convergence of the IHT algorithm, we designed a scheme for expanding the net-work by replicating the existing hidden neurons. Thus we can start training from a small and dense network which is already learned. (3) We implemented this algorithm in GPU. Under small minibatch sizes and large network sizes (e.g., 2000 hidden neurons) it achieves 160 times speedup compared to the RNNLM toolkit in CPU. With larger mini-batch sizes there could be another 10 times speedup, though the convergence rate be-comes an issue in such cases and further effort is needed to address this problem. (4) Without theoretical convergence guarantee of the IHT algorithm in our problem setting, we did an empirical study showing that learning a sparse network does give competitive perplexity in language modeling. In particular, we showed that a sparse network learned in this way can outperform a dense network when the number of effective parameters is kept the same. (5) We gathered performance metric comparing the computational effi-ciency of the matrix operations of interest in both sparse and dense settings. The results suggest that for network sizes which we can train in reasonable time at this moment, it’s hard for sparse matrices to run faster, unless we are allowed to have very sparse networks. Thus for research purposes we may want to focus on using dense matrices, while for en-gineering purposes a more flexible matrix design leveraging the power of dense and sparse matrices might be necessary.

Committee:

Eric Fosler-Lussier, Dr. (Advisor); Mikhail Belkin, Dr. (Committee Member)

Subjects:

Artificial Intelligence; Computer Science

Keywords:

language modeling; recurrent neural networks; sparse recurrent neural networks

Bettaieb, Luc AlexandreA Deep Learning Approach To Coarse Robot Localization
Master of Sciences (Engineering), Case Western Reserve University, 2017, EECS - Electrical Engineering
This thesis explores the use of deep learning for robot localization with applications in re-localizing a mislocalized robot. Seed values for a localization algorithm are assigned based on the interpretation of images. A deep neural network was trained on images acquired in and associated with named regions. In application, the neural net was used to recognize a region based on camera input. By recognizing regions from the camera, the robot can be localized grossly, and subsequently refined with existing techniques. Explorations into different deep neural network topologies and solver types are discussed. A process for gathering training data, training the classifier, and deployment through a robot operating system (ROS) package is provided.

Committee:

Wyatt Newman (Advisor); Murat Cavusoglu (Committee Member); Gregory Lee (Committee Member)

Subjects:

Computer Science; Electrical Engineering; Robotics

Keywords:

robotics; localization; deep learning; neural networks; machine learning; state estimation; robots; robot; robot operating system; ROS; AMCL; monte carlo localization; particle filter; ConvNets; convolutional neural networks

Howard, Shaun MichaelDeep Learning for Sensor Fusion
Master of Sciences (Engineering), Case Western Reserve University, 2017, EECS - Computer and Information Sciences
The use of multiple sensors in modern day vehicular applications is necessary to provide a complete outlook of surroundings for advanced driver assistance systems (ADAS) and automated driving. The fusion of these sensors provides increased certainty in the recognition, localization and prediction of surroundings. A deep learning-based sensor fusion system is proposed to fuse two independent, multi-modal sensor sources. This system is shown to successfully learn the complex capabilities of an existing state-of-the-art sensor fusion system and generalize well to new sensor fusion datasets. It has high precision and recall with minimal confusion after training on several million examples of labeled multi-modal sensor data. It is robust, has a sustainable training time, and has real-time response capabilities on a deep learning PC with a single NVIDIA GeForce GTX 980Ti graphical processing unit (GPU).

Committee:

Wyatt Newman, Dr (Committee Chair); M. Cenk Cavusoglu, Dr (Committee Member); Michael Lewicki, Dr (Committee Member)

Subjects:

Artificial Intelligence; Computer Science

Keywords:

deep learning; sensor fusion; deep neural networks; advanced driver assistance systems; automated driving; multi-stream neural networks; feedforward; multilayer perceptron; recurrent; gated recurrent unit; long-short term memory; camera; radar;

Hope, PriscillaUsing Artificial Neural Networks to Identify Image Spam
Master of Science, University of Akron, 2008, Computer Science

Internet technology has made international communication easy and convenient. This convenience has compelled a number of people to rely on electronic mail for almost all spheres of life – personal, business etc. Scrupulous organizations/individuals have taken undue advantage of this convenience and populate users’ inboxes with unwanted messages making email spam a menace. Even as anti-spam software producers think they have almost solved the problem, spammers come out with new techniques. One such tactic in the spammers’ toolbox comes in the form of image spam – messages that contain little more than a link to an image rendered in an HTML mail reader. The image typically contains the spam message one hopes to avoid, yet it is able to bypass most filters due to the composition and format of these pictures.

This research focuses on identifying these images as spam by using an artificial neural network (ANN), software programs used for recognizing patterns, based on the biological neural networks in our brains. As information propagates through a neural network, it “learns” about the data. A large collection of both spam and non-spam images have being used to train an ANN, and then test the effectiveness of the trained network against an unidentified or already identified set of pictures. This process involves formatting images and adding the desired training values expected by the ANN. Several different ANNS have being trained using different configurations of hidden layers and nodes per layer. A detailed process for preprocessing spam image files is given, followed by a description on how to train an artificial neural network to distinguish between ham and spam. Finally, the trained network is tested against both known and unknown images.

Committee:

Kathy Liszka, PhD (Advisor); Timothy O’Neil (Other); Tim Marguish (Other)

Subjects:

Computer Science

Keywords:

image spam; FANN; artificial neural networks; using artificial neural networks to identify image spam

Sahinoglu, Mehmet MuratDevelopment of a real-time learning scheduler using adaptive critics concepts
Master of Science (MS), Ohio University, 1993, Industrial and Manufacturing Systems Engineering (Engineering)

Manufacturing systems involve human beings, machinery, non-linear dynamics, and a fusion of hierarchical and distributed organizational schemes. The cost-effective control and scheduling of these manufacturing systems require adaptability. Therefore, it is very important to implement competent on-line learning mechanisms that may accomplish balanced and adequate operation of processes with unknown dynamics in the manufacturing environment.

The integration of Dynamic Programming and Artificial Neural Networks has very important characteristics which may provide efficient real-time learning mechanisms for manufacturing. This thesis presents a system that achieves real-time learning using the mentioned integration for manufacturing scheduling. The system is capable of operational mappings. In addition, it utilizes reinforcement signals of the environment (a measure of how desirable the achieved state is taking into consideration the performance criteria) due to the lack of an expert scheduler. Conclusions are drawn and further research issues are discussed.

Committee:

Luis Rabelo (Advisor)

Subjects:

Engineering, Industrial

Keywords:

Real-Time Learning Scheduler; Adaptive Critics; Artificial Neural Networks; Dynamic Programming

Egilmez, GokhanRoad Safety Assessment of U.S. States: A Joint Frontier and Neural Network Modeling Approach
Master of Science (MS), Ohio University, 2013, Civil Engineering (Engineering and Technology)
In this thesis, road safety assessment and prediction modeling for U.S. states fatal crashes are addressed. In the first part, a DEA-based Malmquist Index model was developed to assess the relative efficiency and productivity of U.S. states in decreasing the number of road fatalities. Even though the national trend in fatal crashes has reached to the lowest level since 1949 (Traffic Safety Annual Assessment Highlights, 2010), a state-by-state analysis and comparison has not been studied considering other characteristics of the holistic national road safety assessment problem in any work in the literature or organizational reports. The single output, fatal crashes, and five inputs were aggregated into single road safety score and utilized in the DEA-based Malmquist Index mathematical model. The period of 2002-2008 was considered due to data availability for the inputs and the output considered. According to the results, there is a slight negative productivity (an average of -0.2 percent productivity) observed in the U.S. on minimizing the number of fatal crashes along with an average of 2.1 percent efficiency decline and 1.8 percent technological improvement. The productivity in reducing the fatal crashes can only be attributed to the technological growth since there is a negative efficiency growth is occurred. It can be concluded that even though there is a declining trend observed in the fatality rates, the efficiency of states in utilizing societal and economical resources towards the goal of zero fatality is not still efficient. In the second part, a nonparametric prediction model, Artificial Neural Network, was developed to assist policy makers in minimizing fatal crashes across the United States. Seven input variables from four safety performance input domains while fatal crashes was utilized as the single output variable for the scope of the research. Artificial Neural Networks (ANN) was utilized and the best neural network model was developed out of 1000 networks. The proposed neural network model predicted data with 84 percent coefficient of determination. In addition, developed ANN model was benchmarked with a multiple linear regression model and outperformed in all performance metrics including r, R2 and the standard error of estimate. A sensitivity analysis was also conducted and the results indicated that road length, vehicle miles traveled, and safety expenditures were the top three input variables on fatal crashes. In conclusion, more effective policy making towards increasing safety belt usage and better utilization of safety expenditures to improve road condition are derived as the key areas to focus on for state highway safety agencies from the scope of current research. This research also reveals the significance of the relationship between the four input domains and fatal crashes for the United States from a holistic perspective and offers a robust nonparametric model to policy makers for the prediction of fatal crashes.

Committee:

Deborah McAvoy, Ph.D. (Advisor); Byung-Cheol Kim, Ph.D. (Committee Member); Ken Walsh, Ph.D. (Committee Member); M. Khurrum S. Bhutta, Ph.D. (Committee Member)

Subjects:

Civil Engineering; Industrial Engineering; Transportation

Keywords:

Road Safety Assessment; Benchmarking; Data Envelopment Analysis; Malmquist Productivity Index; Nonparametric Predictive Modeling; Artificial Neural Networks; Machine Learning; US States

Han, BingACCELERATION OF SPIKING NEURAL NETWORK ON GENERAL PURPOSE GRAPHICS PROCESSORS
Master of Science (M.S.), University of Dayton, 2010, Electrical Engineering
There is currently a strong push in the research community to develop biological scale implementations of neuron based vision models. Systems at this scale are computationally demanding and have generally utilized more accurate neuron models, such as the Izhikevich and Hodgkin-Huxley models, in favor of the more popular integrate and fire model. This thesis examines the feasibility of using GPGPUs for accelerating a spiking neural network based character recognition network to enable large scale neural systems. Two versions of the network utilizing the Izhikevich and Hodgkin-Huxley models are implemented. Three NVIDIA GPGPU platforms and one GPGPU cluster were examined. These include the GeForce 9800 GX2, the Tesla C1060, the Tesla S1070 platforms, and the 32-node Tesla S1070 GPGPU cluster. Our results show that the GPGPUs can provide significant speedups over conventional processors. In particular, the fastest GPGPU utilized, the Tesla S1070, provided speedups of 5.6 and 84.4 time over highly optimized implementations on the fastest CPU tested, a quad core 2.67 GHz Xeon processor, for the Izhikevich and Hodgkin Huxley models respectively. The CPU implementation utilized all four cores and the vector data parallelism offered by the processor. The results indicate that GPGPUs are well suited for this application domain. A large portion of the results presented in this thesis have been published in the April 2010 issue of Applied Optics [1].

Committee:

Tarek Taha (Committee Chair); John Loomis (Committee Member); Balster Eric (Committee Member)

Subjects:

Electrical Engineering

Keywords:

Spiking neural networks; GPGPU

Holland, William S.Development of an Indoor Real-time Localization System Using Passive RFID Tags and Artificial Neural Networks
Master of Science (MS), Ohio University, 2009, Industrial and Systems Engineering (Engineering and Technology)
Radio frequency identification (RFID) technology is used for inventory and asset tracking because of its accuracy and speed. Currently, RFID tracking systems are being used to identify and locate tagged objects in indoor environments. In this research, received signal strength indication (RSSI) values are collected from off-the-shelf passive RFID readers and antennas to be used in conjunction with an artificial neural network (ANN) to create a localization algorithm for two-dimensional location estimation with a single tag. The aim of this research is to create a highly accurate real-time location tracking system to be used in a room with objects that create RF interference. Multiple linear regression is used as a benchmark method for comparison with artificial neural networks.

Committee:

Gary Weckman, PhD (Advisor); Kevin Berisso, PhD (Committee Member); Diana Schwerha, PhD (Committee Member); Andrew Snow, PhD (Committee Member)

Subjects:

Artificial Intelligence; Engineering; Industrial Engineering; Systems Design

Keywords:

RFID; RSSI; artificial neural networks; location system

Lakshminarayanan, SriramAn Integrated Stock Market Forecasting Model Using Neural Networks
Master of Science (MS), Ohio University, 2005, Industrial and Manufacturing Systems Engineering (Engineering)

This thesis focuses on the development of a stock market forecasting model based on an Artificial Neural Network architecture. This study constructs a hybrid model utilizing various technical indicators, Elliott’s wave theory, sensitivity analysis and fuzzy logic. Initially a baseline network is constructed based on available literature. The baseline model is then improved by applying several useful information domains to the different models. Optimizations of the Neural Network models are performed by augmenting the network with useful information at every stage.

Committee:

Gary Weckman (Advisor)

Subjects:

Engineering, Industrial

Keywords:

Neural Networks; Forecasting; Stock Markets

Anderson, Jerone S.A Study of Nutrient Dynamics in Old Woman Creek Using Artificial Neural Networks and Bayesian Belief Networks
Master of Science (MS), Ohio University, 2009, Industrial and Systems Engineering (Engineering and Technology)
The Old Woman Creek National Estuary is studied in this project to evaluate effective modelling techniques for predicting Net Ecosystem Metabolism (NEM). NEM is modelled using artificial neural networks, Bayesian belief networks, and a hybrid model. A variety of data preprocessing techniques are considered prior to model development. The effects of discretization on model development are considered and discrete data is ultimately used to produce models which classify NEM into three ranges based on inputs with information significance. Artificial neural networks are found to be the most accurate for classification while Bayesian belief networks are found to provide a better framework for dynamically predicting NEM as inputs are changed.

Committee:

Gary R. Weckman, PhD (Advisor); David Millie, PhD (Committee Member); Kevin Berisso, PhD (Committee Member); Diana Schwerha, PhD (Committee Member)

Subjects:

Ecology; Engineering; Environmental Engineering; Industrial Engineering

Keywords:

BBN; ANN; ecology; NEM; Bayesian Belief Networks; Artificial Neural Networks; computer modelling

Brown, Marvin LaneThe Impact of Data Imputation Methodologies on Knowledge Discovery
Doctor of Business Administration, Cleveland State University, 2008, Nance College of Business Administration

The purpose of this research is to investigate the impact of Data Imputation Methodologies that are employed when a specific Data Mining algorithm is utilized within a KDD (Knowledge Discovery in Databases) process. This study will employ certain Knowledge Discovery processes that are widely accepted in both the academic and commercial worlds. Several Knowledge Discovery models will be developed utilizing secondary data containing known correct values. Tests will be conducted on the secondary data both before and after storing data instances with known results and then identifying imprecise data values. One of the integral stages in the accomplishment of successful Knowledge Discovery is the Data Mining phase. The actual Data Mining process deals significantly with prediction, estimation, classification, pattern recognition and the development of association rules. Neural Networks are the most commonly selected tools for Data Mining classification and prediction. Neural Networks employ various types of Transfer Functions when outputting data. The most commonly employed Transfer Function is the s-Sigmoid Function. Various Knowledge Discovery Models from various research and business disciplines were tested using this framework.

However, missing and inconsistent data has been pervasive problems in the history of data analysis since the origin of data collection. Due to advancements in the capacities of data storage and the proliferation of computer software, more historical data is being collected and analyzed today than ever before. The issue of missing data must be addressed, since ignoring this problem can introduce bias into the models being evaluated and lead to inaccurate data mining conclusions. The objective of this research is to address the impact of Missing Data and Data Imputation on the Data Mining phase of Knowledge Discovery when Neural Networks are utilized when employing an s-Sigmoid Transfer function, and are confronted with Missing Data and Data Imputation methodologies.

Committee:

Chien-Hua (Mike) Lin, Phd (Committee Chair); Adam Fadlalla, Phd (Committee Member); Walter Rom, Phd (Committee Member); John Kros, Phd (Committee Member); Marc Lynn, Phd (Advisor)

Subjects:

Business Education; Computer Science

Keywords:

Data Mining; Knowledge Discovery; Data Imputation; Neural Networks; Transfer Functions; Sigmoid

Perumal, SubramoniamStability and Switchability in Recurrent Neural Networks
MS, University of Cincinnati, 2008, Engineering : Computer Science

Artificial Neural Networks (ANNs) are being extensively researched for their wide range of applications. Among the most important is the ability of a type of ANNs—recurrent attractor networks—to work as associative memories. The most common type of ANN used for associative memory is the Hopfield network, which is a fully connected network with symmetric connections. There have been numerous attempts to improve the capacity and recall quality of recurrent networks, with the focus primarily on the stability of the stored attractors, and the network's convergence properties. However, the ability of a recurrent attractor network to switch between attractors is also an interesting property, if it can be harnessed for use. Such switching can be useful as a model of switching between context-dependent functional networks thought to underlie cognitive processing.

In this thesis, we design and develop a stable-yet-switchable (SyS) network model which provides an interesting combination of stability and switchability. The network is stable under random perturbations, but highly sensitive to specific targeted perturbations which cause it to switch attractors. Such functionality has previously been reported in networks with scale-free (SF) connectivity. We introduce networks with two regions: A densely connected core region, and a sparsely connected and larger periphery. We show that these core-periphery (CP) networks are better for providing a combination of stability and targeted switching than scale-free networks. We develop and validate a specific approach to switching between attractors in a targeted way. The CP and SF models are also compared with each other and with randomly connected homogeneous networks.

Committee:

Dr. Ali Minai (Advisor); Dr. Raj Bhatnagar (Committee Member); Dr. Anca Ralescu (Committee Member)

Subjects:

Computer Science; Engineering

Keywords:

recurrent neural networks; core-periphery networks; switchability; switching between attractors; stability and switchability

Sneath, Evan BArtificial neural network training for semi-autonomous robotic surgery applications
MS, University of Cincinnati, 2014, Engineering and Applied Science: Computer Engineering
As telesurgical robots become more frequently used in surgical operating rooms, emphasis is shifting from human-controlled robotics to semi- or full automaticity. Safe and efficient methods of training and execution during an automated surgical task are required for real-world success. The approach of path generation using artificial neural networks allows for an effective and scalable solution for the supervised learning and real-time performance of a surgical procedure. This study makes use of long short-term memory (LSTM) recurrent neural networks (RNNs) in conjunction with the Evolino learning algorithm for tooltip path optimization. The RNNgenerated path is trained from human-performed procedures in a simulated testing environment. Changes in movement of path markers are accounted for by adjusting the tooltip acceleration with respect to target markers along the path. Results include smooth generated paths successfully meeting test procedure requirements of accuracy and speed in environments with both static and dynamic marker configurations.

Committee:

Fred Beyette, Ph.D. (Committee Chair); Ali Minai, Ph.D. (Committee Member); Grant Schaffner, Ph.D. (Committee Member)

Subjects:

Computer Engineering

Keywords:

telesurgery;robotics;artificial neural networks;evolino

Muralidharan Nair, MithunStatistical Leakage Estimation Using Artificial Neural Networks
MS, University of Cincinnati, 2014, Engineering and Applied Science: Computer Engineering
Present day integrated circuit designs have become very densely packed with smaller devices. The scaling down of technology has increased the significance of modeling the effect of process variations. Increasing leakage power consumption is another factor that the circuit designers are concerned about, in the smaller technology nodes. This continuous reduction in size of the devices has made engineers to give high importance to the effect of process variations in these designs. The effect of variation can be very drastic when variations affect the functionality of the chip. There could be a finite probability that the chip is functional in the presence of variations but does not meet performance and/or power consumption requirements. The modern day chip designing is very much oriented towards high performance and low power designs. In this scenario, an adverse effect from process variations can prove counter productive for the designer. A good design methodology should be able to predict and address these adverse effects at early design stages. So it is the need of the hour to have techniques that could model the effect of process variations from synthesis to post routing stages. In this thesis we propose a methodology that could accurately estimate the leakage in the presence of variations after synthesis stage. Since we are addressing the issue at pre-layout stage, we have given importance to the variations in device dimensions and threshold voltage. The methodology starts from the RTL description of a design. This design is synthesized to a netlist of Standard Cells. We have used all the standard cell definitions and characterized power and timing values from Synopys 90nm EDK. This netlist is used for our experiments. The core of this method is the artificial neural network models for standard cell leakage. These models are generated for a wide range of cells in the standard cell library and in turn are used in a tool that produces a Statistical Leakage estimate in the presence of process variations. The standard methodology for statistical leakage analysis in the presence of variations is Monte Carlo simulations in HSPICE. The conventional Statistical Leakage estimation is done using Monte Carlo simulations for smaller circuits. For larger benchmarks one of the method used for Statistical Leakage estimation is Wilkinson's approach. This method is applied on standard circuit benchmarks and a comparison has been made with the conventional Statistical Leakage estimation methodologies.

Committee:

Ranganadha Vemuri, Ph.D. (Committee Chair); Wen Ben Jone, Ph.D. (Committee Member); Carla Purdy, Ph.D. (Committee Member)

Subjects:

Computer Engineering

Keywords:

Process Variations;Leakage Power;Statistical Leakage;Artificial Neural Networks

Halsey, Phillip A.The Nature of Modality and Learning Task: Unsupervised Learning of Auditory Categories
Doctor of Philosophy (PhD), Ohio University, 2015, Experimental Psychology (Arts and Sciences)
Categorization and concept-learning has a long-standing influence on the field of psychology because the notions of concept-learning are key to how individuals learn. Central to this idea is; how do we categorize stimuli that vary according to different dimensions? How do we categorize stimuli under different conditions? How do we store these categorizes as mental representations? And does the modality of the stimuli affect our construction of a mental concept, and to what extent does this affect categorization behavior? To partially answer this last question, it has been determined that the modality of a stimulus does influence categorization behavior but the extent of this is unknown. The current dissertation explores the manner in which stimulus modality, relationships between stimulus dimensions, and learning method affects categorization behavior. Two experiments are conducted in order to examine the auditory dimensions individuals attend to when making comparisons, and how individuals spontaneously categorize auditory stimuli based on the attended dimensions. Participant’s data was then examined according to three models of unsupervised learning: the simplicity model, SUSTAIN, and GISTM.

Committee:

Ronaldo Vigo (Advisor); Steve Evans (Committee Member); Keith Markman (Committee Member); Robert Briscoe (Committee Member); Mark Phillips (Committee Member)

Subjects:

Cognitive Psychology; Psychology

Keywords:

unsupervised learning; audio; concept learning; mathematical modeling; invariance; neural networks

Macmann, OwenPerforming Diagnostics & Prognostics On Simulated Engine Failures Using Neural Networks
MS, University of Cincinnati, 2016, Engineering and Applied Science: Aerospace Engineering
Good prognostic health management (PHM) solutions for jet engines remain elusive, owing partially to lack of run-to-failure data sets. A good PHM solution has the potential to improve on unscheduled maintenance by offering an accurate, real-time estimation of the engine’s current health state. Aero-engine simulations allow for generation of simulated data invaluable for data-driven PHM solutions. Simulated data can characteristically represent propagation of faults in an engine over time and present the results of that fault propagation in terms of realistically acquirable sensor data. A method of data set generation for jet engine degradation that incorporates multiple faults is described. The generated data sets can be used for training a combined diagnostic/prognostic solution. This work proposes a neural network-based prognostic system that uses diagnostic evaluations as additional tag data for a prognostic analysis. Self-organized maps are used to classify data. The classifications are added to the data as an additional input for a neural network designed to predict remaining usable life. The method exhibits totally autonomous learning of data and produces improvements over approaches that do not pre-classify data.

Committee:

Kelly| Cohen, Ph.D. (Committee Chair); Al Behbahani, Ph.D. (Committee Member); Kristin Yvonne Rozier|, Ph.D. (Committee Member); Bruce Walke, Sc.D. (Committee Member)

Subjects:

Aerospace Materials

Keywords:

prognostics;diagnostics;neural networks;engine health management;degradation simulation;self-organized maps

Regmi, Hem KantaA Real-Time Computational Decision Support System for Compounded Sterile Preparations using Image Processing and Artificial Neural Networks
Master of Science, University of Toledo, 2016, Electrical Engineering
The purpose of this research is to design a computational decision support system (DSS) for compounded sterile preparations (CSP). Error-free compounding is dependent on the proper selection of components and adherence to procedure during compounding. A material selection system (MSS) based on a graphical user interface (GUI), coupled with a barcode scanner and back-end database, has been developed and tested for proper selection of items involving three different medication orders (MO). A video processing system (VPS) has been implemented in MATLAB that evaluates the live video feed from the compounding hood to monitor the compounding procedure when compounding the MO’s. Surf detection is used to detect and locate compounding items placed in the hood. Various algorithms have been developed and tested to enhance the accuracy and robustness of the VPS. The Decision Support System (DSS) is further improved with integration of another digital camera to ensure that correct volume of medicine with appropriate syringe is performed during the whole compounding process. The template matching and SURF object detection application on the digital image of the syringe, along with minimum distance classifier and artificial neural networks (ANNs) on the previously collected data from several experimental observations, were explored in classification and volume measurement of a syringe. The MSS was tested for all items used in compounding the MO’s and performed error-free. The VPS evolved to VPS.03 from VPS.01 and VPS.02. The greatest accuracy and ability for real-time realization were seen in VPS.03. All deliberate mistakes made when compounding the tested medication orders were captured by VPS.03. Leur-lock syringes of different sizes from 1 mL to 30 mL were tested, and an accuracy of 95+ % was obtained with very high precision. The new computational decision support system facilitates error-free selection of components and is able to monitor and evaluate the compounding process and correct volume measurement in real time. The platform may be used in CSP compounding rooms to audit techniques and procedures as well as in training or educational settings.

Committee:

Vijay Devabhaktuni, Dr. (Committee Chair); Jerry Nesamony, Dr. (Committee Co-Chair); Devinder Kaur, Dr. (Committee Member); Ezzatollah Salari, Dr. (Committee Member)

Subjects:

Electrical Engineering

Keywords:

Compounding Sterile Preparations, Graphical User Interface, Artificial Neural Networks, Image Processing, Video Processing, Decision Support System, Medication Order, Object Detection, Correlation Calculation, Connected Component Analysis

Next Page