A new and unique directional weighted error function was introduced into the backpropagation algorithm used in Artificial Neural Networks (ANNs) for applications where yes or no decisions are made on the output. A continuous error function based on a weighted curve is suggested for use in the backpropagation algorithm in an effort to increase the number of correct decisions. Results were compared to the standard and weighted error methods. A higher number of correct decisions were made with the new method. / Thesis (M.S.)--Wichita State University, College of Engineering, Dept. of Electrical and Computer Engineering. / "July 2006." / Includes bibliographic references (leaves 36-37).

A new methodology was developed for flow regime identification in pipes. The method utilizes the pattern recognition abilities of Artificial Neural Networks and the unprocessed time series of a system-monitoring-signal. The methodology was tested with synthetic data from a conceptual system, liquid level indicating Capacitance signals from a Horizontal flow system and with a pressure difference signal from a S-shape riser. The results showed that the signals that were generated for the conceptual system had all their patterns identified correctly with no errors what so ever. The patterns for the Horizontal flow system were also classified very well with a few errors recorded due to original misclassifications of the data. The misclassifications were mainly due to subjectivity and due to signals that belonged to transition regions, hence a single label for them was not adequate. Finally the results for the S-shape riser showed also good agreement with the visual observations and the few errors that were identified were again due to original misclassifications but also to the lack of long enough time series for some flow cases and the availability of less flow cases for some flow regimes than others. In general the methodology proved to be successful and there were a number of advantages identified for this neural network methodology in comparison to other ones and especially the feature extraction methods. These advantages were: Faster identification of changes to the condition of the system, inexpensive suitable for a variety of pipeline geometries and more powerful on the flow regime identification, even for transitional cases.

An electric power system consists of the generating stations, the transmission lines, and the distribution systems. Transmission lines are the connecting links between the generating stations and the distribution systems. With the rapid growth of economy and technology, the demand for large blocks of power, power quality and increased reliability suggested the interconnection of neighboring systems. Transmission lines are elements of a network which connects the generating plants to the distribution systems, and could extend hundreds of miles . Because of the long distances traversed by transmission lines over open area, they tend to fade by natural and artificial calamity imposed on the power system. It maybe easy to discover the fault with sufficient information in the populous region. When fault occurs in the remote region, it is difficult to identify the outage location. An efficient and reliable technique is thus desirable to resolve the problem. This dissertation presents the fault location for high voltage lines with Artificial Neural Network( ANN ) method. Beside the fault location, this research also improve the problem further by considering the fault resistance. The fault resistance may not remain the same due to the variation of environmental factors. The fault location may involve errors owing to the fault resistance. An algorithms has been developed in this dissertation to calculate fault resistance and revise the ANN training data for three-phase fault, double line-to-ground fault, single line-to-ground fault, and line-to-line fault. To verify the effectiveness of the method, practical transmission lines were used for tests. The results proved that the method could be used to identify the fault location effectively and help dispatchers determine a reference distance.

­^¤å´£­n¡G A utility function model of individual credit card holder based on their spending behavior is constructed in this research. An accumulation of the individual utility of three different risk attitudes of cardholders may be useful for promoting the profits of credit card business for banks. Due to the privacy of cardholders and the lack of real data, a questionnaire sampling is used to collect data for this study. A result of this experimental study indicates that credit card holders with a different sex, age, level of education, asset condition, seniority, and occupation have different risk tendency. Based on 249 effective samples in this research, credit card holders who belong to females, teenagers, relatively low educated, without real estate, middle seniority, and relatively volatile occupation are more risk seeking. Relatively risk seeking credit card holders have the tendency to make use of their revolving credit and to borrow cash or to buy financial products with their credit cards. For those with three different risk attitudes, their default of credit card loans are not significantly different. The finding indicates relatively risk seeking cardholders may contribute more profits to the credit card business for banks. A risk attitude classification model built by artificial neural network has also been developed. The model may assist banks' administrators using their applicants' demographics to distinguish their risk attitude for approving an appropriate credit limit for a cardholder's expenditure to promote the total credit card profit for banks.

A Microsoft Windows program is designed to implement a text independent speaker recognition system in numerous speaker cases based on Mel-Cepstrum and hierarchical tree classifier and binary vector quantization. Experimental result show that the accuracy is barely affected by increasing population sizes. And the speed of recognizing is fast than traditional methods.

The air quality in Kaohsiung and Ping-Dong district is the worst in Taiwan. The air pollution episodes in Kaohsiung are attributed to high concentrations of PM10 and O3. Among them, over half of the episodes result from PM10. In addition to Pollutant Standards Index (PSI), atmospheric visibility is also an indicator of ambient air quality. Citizens always complain about the impairment of visibility because it can be visualized directly. Visibility is closely correlated to both air pollutants and meteorological condition. Extinction of visible light by fine particles is the major reason for visibility impairment. In this study, an artificial neural network was applied to predict the concentration of PM10 and atmospheric visibility. The objectives of this study were to investigate the effects of meteorological factor and air pollutants on visibility and to apply artificial neural network to predict the concentration of PM10 and atmospheric visibility. The measured PM10 data were divided into two parts (i.e. summer and winter, ) to understand whether different season affect the prediction of PM10 concentration. The modeling results showed that the optimum input variables included the PM10 concentration, atmospheric pressure, surface radiation, relative humidity, atmospheric temperature, and cloud condition. The network outputs showed high correlation with measured PM10 concentration (R=0.876) in the whole-year set. Furthermore, the prediction of summer set also showed high correlation with measured PM10 concentration (R=0.753). The winter set demonstrated the worse prediction among three sets, and showed medium correlation with measured PM10 concentration (R=0.553). The visibility network test was conducted by two stages. The first stage (set-1~set-3) showed that relative humidity, atmospheric temperature, and cloud condition were the most important meteorological factors, while PM10, O3, and NO3 were the most important air pollutants on the prediction of atmospheric visibility. The prediction of set-1 considering only meteorological factors was the worst (R=0.586), while set-3 was the best and showed medium correlation with measured atmospheric visibility (R=0.633). The second stage (set-4 and set-5) increased the hidden neuron numbers and input variables, and added atmospheric visibility in the input variables. Although the correlation coefficients between predicted and measured data did not increase, the prediction of atmospheric visibility had significant improvement. Finally, a short-term prediction of PM10 and atmospheric visibility was conducted and validated by the level of PSI values and atmospheric visibility. Prediction results showed that the accuracy of PM10 prediction was 76.9%, while the prediction of atmospheric visibility by set-3 network demonstrated an accuracy of 76.9%. Moreover, no significant difference of prediction was detected by using either three-level or five-level visibility systems.

Secret sharing is a fundamental notion for secure cryptographic design. In a secret sharing scheme, a set of participants shares a secret among them such that only pre-specified subsets of these shares can get together to recover the secret. This dissertation introduces a neural network approach to solve the problem of secret sharing for any given access structure. Other approaches have been used to solve this problem. However, the yet known approaches result in exponential increase in the amount of data that every participant need to keep. This amount is measured by the secret sharing scheme information rate. This work is intended to solve the problem with better information rate.

In this thesis, we study the least trimmed squares artificial neural networks (LTS-ANNs), which are generalization of the least trimmed squares (LTS) estimators frequently used in robust linear parametric regression problems to nonparametric artificial neural networks (ANNs) used for nonlinear regression problems. Two training algorithms are proposed in this thesis. The first algorithm is the incremental gradient descent algorithm. In order to speed up the convergence, the second training algorithm is proposed based on recursive least squares (RLS). Three illustrative examples are provided to test the performances of robustness against outliers for the classical ANNs and the LTS-ANNs. Simulation results show that upon proper selection of the trimming constant of the learning machines, LTS-ANNs are quite robust against outliers compared with the classical ANNs.

<p>Today one of the greatest aims within the area of bioinformatics is to gain a complete understanding of the functionality of genes and the systems behind gene regulation. Regulatory relationships among genes seem to be of a complex nature since transcriptional control is the result of complex networks interpreting a variety of inputs. It is therefore essential to develop analytical tools detecting complex genetic relationships.</p><p>This project examines the possibility of the data mining technique artificial neural network (ANN) detecting regulatory relationships between genes. As an initial step for finding regulatory relationships with the help of ANN the goal of this project is to train an ANN to predict the expression of an individual gene. The genes predicted are the nuclear receptor PPAR-g and the insulin receptor. Predictions of the two target genes respectively were made using different datasets of gene expression data as input for the ANN. The results of the predictions of PPAR-g indicate that it is not possible to predict the expression of PPAR-g under the circumstances for this experiment. The results of the predictions of the insulin receptor indicate that it is not possible to discard using ANN for predicting the gene expression of an individual gene.</p>

This report examines the trends in sick leave usage among nurses in a hospital and aims at creating a forecasting model to predict sick leave usage on a weekly basis using the concept of artificial neural networks (ANN). The data used for the research includes the absenteeism (sick leave) reports for 3 years at a hospital. The analysis shows that there are certain factors that lead to a rise or fall in the weekly sick leave usage. The ANN model tries to capture the effect of these factors and forecasts the sick leave usage for a 1 year horizon based on what it has learned from the behavior of the historical data from the previous 2 years. The various parameters of the model are determined and the model is constructed and tested for its forecasting ability. / text