Lee A. Belfore

A syntactic methodology for automatic diagnosis by analysis of continuous time measurements using hierarchical signal representations

In this paper, we present a methodology for automatic diagnosis of systems characterized by continuous signals. For each condition considered, the methodology requires the development of an alphabet of signal primitives, and a set of hierarchical fuzzy automatons (HFAs). Each alphabet is adaptively obtained by training an adaptive resonance theory (ART2) architecture with signal segments from a particular condition. Then, the original signal is transformed into a string of vectors of primitives, where each vector of primitives replaces a signal segment in the original signal. The string, in turn, is presented to the HFA characterizing that particular condition. Each set of HFA consists of a main automaton identifying the entire signal, and several sub-automata each identifying a particular significant structure in the signal. A transition in the main automaton occurs (i.e., the main automaton moves from one state to another) if the corresponding subautomaton recognizes a token where a token is a portion of the string of vectors of signal primitives with a significant structure. The fuzziness in automaton operation adds flexibility to the operation of the automaton, enabling the processing of imperfect input, allowing for toleration measurement noise and other ambiguities. The methodology is applied to the problem of automatic electrocardiogram diagnosis.

Modeling of fault tolerance in neural networks

The authors present an analytical technique for assessing the fault tolerance of neural networks. The basis of the technique is developed through an analogy with magnetic spin systems using statistical mechanics. It is shown that neural networks can be analyzed using statistical mechanics. Simulated results are compared with analytical results, showing that the analytical model does indeed conform to the simulation model. The primary example presented is an associative memory.<<ETX>>

An interactive land use VRML application (ILUVA) with servlet assist

We summarize progress achieved on an interactive land use VRML application (ILUVA) with servlet assist. The purpose of this application is to enable one to take a virtual land area and add buildings, roadways, landscaping and other features. The application is implemented entirely using standard Web based technologies to allow fairly universal accessibility. The Virtual Reality Modeling Language (VRML) is a programming language that describes three dimensional objects and defines interactions associated with these objects. In this work, we show how the interactive capabilities can be expanded by employing Java servlets for recording user actions and for restoring prior sessions. The Java servlets offer several powerful capabilities including enabling logging permanent records of user sessions, retrieval of prior sessions, and dynamically generated VRML.

The analysis of the faulty behavior of synchronous neural networks

A means for analyzing the faulty behavior of neural networks is presented. Using an analogy between statistical physics and neural networks, a method for assessing the performance of a synchronous neural network model in the presence of faults is developed. Analytical predictions are computed using the statistical physics analogy and compared with the simulated behavior for two neuron models. An example of the analytical technique applied to an autoassociative memory is described. >

Using a computer simulation for teaching communication skills: A blinded multisite mixed methods randomized controlled trial.

OBJECTIVES To assess advanced communication skills among second-year medical students exposed either to a computer simulation (MPathic-VR) featuring virtual humans, or to a multimedia computer-based learning module, and to understand each groups experiences and learning preferences. METHODS A single-blinded, mixed methods, randomized, multisite trial compared MPathic-VR (N=210) to computer-based learning (N=211). Primary outcomes: communication scores during repeat interactions with MPathic-VRs intercultural and interprofessional communication scenarios and scores on a subsequent advanced communication skills objective structured clinical examination (OSCE). Multivariate analysis of variance was used to compare outcomes. SECONDARY OUTCOMES student attitude surveys and qualitative assessments of their experiences with MPathic-VR or computer-based learning. RESULTS MPathic-VR-trained students improved their intercultural and interprofessional communication performance between their first and second interactions with each scenario. They also achieved significantly higher composite scores on the OSCE than computer-based learning-trained students. Attitudes and experiences were more positive among students trained with MPathic-VR, who valued its providing immediate feedback, teaching nonverbal communication skills, and preparing them for emotion-charged patient encounters. CONCLUSIONS MPathic-VR was effective in training advanced communication skills and in enabling knowledge transfer into a more realistic clinical situation. PRACTICE IMPLICATIONS MPathic-VRs virtual human simulation offers an effective and engaging means of advanced communication training.

VRML for urban visualization

A virtual reality modeling language (VRML) based application has been developed as a marketing tool for a commercial park. VRML is a new Web based technology for specifying and delivering three-dimensional interactive visualizations over the Internet through a Web browser. As a part of its definition, VRML includes primitives that specify geometries, sense different conditions in in the visualization, and allow custom definition of methods. Geometries and conditions may be linked so that the geometries can be modified or added interactively. The visualization features simple operation, an extensive menu structure, dynamic creation of objects, and an arbitration scheme.

Multiuser extensions to the Interactive Land Use VRML Application (ILUVA)

Virtual reality simulations implemented in the Virtual Reality Modeling Language (VRML) provide the ability to create Web-based simulation environments with the mark of realism provided by three-dimensional representations. The Interactive Land Use VRML Application (ILUVA) enables users to perform simple site planning by creating building sites and then populating them with buildings, roadways, landscaping, and etc. We describe multiuser extensions that integrate a database interface to allow multiple users to use and save past sessions. The database interface is implemented using servlets and the JDBC provided in the Java core.

Applying hierarchical fuzzy automatons to automatic diagnosis

Hierarchical fuzzy automatons (HFAs) are employed to perform automatic diagnosis on a signal represented as set of discrete time measurements. The HFA incorporates two levels of hierarchy with the lower level identifying structures within the signal and the top level integrating the results from lower level automatons. An adaptive resonance theory (ART) artificial neural network (ANN) is used to determine input tokens and to tokenize the input. The tokens generated by the ANN are given fuzzy memberships using information derived from the state of the ANN. In addition, a general methodology is presented for constructing HFAs. HFAs are applied to the problem of determining whether an ECG recording is normal or shows atrial fibrillation.

Neurogenetic characterization of fault tolerant switched reluctance motor drives

This paper presents the results of a study on the feasibility of using artificial neural networks (ANNs) and genetic algorithms (GAs) to predict the performance characteristics of faulted switched reluctance motor (SRRI) drive systems. In this work, the ANNs are applied for their well known interpolation capabilities for highly nonlinear systems. In addition, the GAs are employed for their ability to search a complex structural and parametric space as necessary to find good ANN solutions. Also, an integrated finite elements/state space modeling approach is used to generate training data sets for the SRM drive system. Furthermore, the results are compared to test data.

Modeling faulted switched reluctance motors using evolutionary neural networks

The work presented examines the feasibility of using artificial neural networks (ANNs) and evolutionary algorithms (EAs) to model fault free and faulted switched reluctance motor (SRM) drive systems. SRMs are capable of functioning despite the presence of faults. Faults impart transient changes to machine inductances in a manner that is difficult to model analytically. After this transient period, SRMs are capable of functioning at a reduced level of performance. ANNs are applied for their well known interpolation capabilities for highly nonlinear systems. EAs are employed for their ability to search a complex structural and parametric space as necessary to find good ANN solutions. In this paper, the ANN structure and training regimen are described for application to an example SRM drive system under normal and abnormal operating conditions.<<ETX>>