Volume2, Number2, 2001


 

 

Title:

The Application of Artificial Neural Networks in Knowledge-Based Information Systems

Author:

Ming Zhang≠, Rex E. Gantenbein, Sung Y. Shin, Chih-Cheng Hung

Abstract:

Artificial neural networks are a novel technique in nonlinear mapping domains that can be applied in a variety of knowledge-based information systems.Specifically, feed-forward artificial neural networks and artificial neural network group-based adaptive (GAT) trees for weather forecasting and face identification systems were investigated in this study.Results from these studies show that neural network techniques can support intelligent information systems.

 

 

Title:

Framework for Network Management Architectures with Distributed Software Component

Author:

Haeng-Kon Kim

Abstract:

High speed communication network take aim at the support of various communication network management and multimedia services based on distributed processing environments. In this paper, we present our experiences with the NDSC (Network Domain Software Component) framework and development environment. The NDSC framework supports component based distributed network management architecture such as TINA (Telecommunications Information Network Architecture). Within the component development environment, components can be developed and evaluated. The NDSC can support both dynamic and static operational interfaces. Also, there can exist multiple instances of the same operational interface. Furthermore, components can be grouped together to form compound components. Through a common control and configuration interface, the components can be configured with regard to events, properties, operational interfaces, life cycle, and composition. We apply this framework to a TINA based services and network management system with component.

 

 

Title:

Partially Opening the Black Box: An ANN with Inspectable, Hidden Layers

Author:

David Primeaux

Abstract:

An artificial neural network (ANN) with a single hidden layer can approximate any computable function.This paper suggests, however, that for some applications, an ANN architecture having multiple hidden layers is appropriate.The ANN discussed here can be made to exhibit pass/fail behavior for the conservation task.Because this ANNís multiple hidden layers are both inspectable and representational, they permit reasonable and useful interpretation of the ANNís computational behavior.

 

 

Title:

High Order Object Oriented Modeling Technique For Structured Object-Oriented Analysis

Author:

Xiaoqing Frank Liu, Lijun Dong, Hungwen Lin

Abstract:

In order to reduce the difficulty in migrating from traditional paradigm into object-oriented paradigm, it is highly desirable to integrate object-oriented modeling with structured analysis seamlessly. Existing approaches suffer from two major problems in this regard. One is a lack of a well-defined processes and mechanisms for structured development of component models consistently based on hierarchical decomposition. Objects are often analyzed at only a single level of abstraction while functional requirements and dynamic behavior are analyzed at multiple levels of abstraction. Another is a lack of well-defined processes and guidelines for integration of different component models. High Order Object-Oriented Modeling Technique (HOOMT) helps develop object, functional, and dynamic models hierarchically according to their abstraction levels. Structural, functional, and dynamic properties of objects at a higher abstraction level can be analyzed based on those of objects at lower abstraction levels. It currently consists of High Order Object Model, Hierarchical Object Information Flow Model, and Hierarchical State Transition Model. HOOMT eliminates incompatibility between a flat object model in which all modeling elements are analyzed at a single level of abstraction, and hierarchical functional and dynamic models, in which modeling elements are analyzed at multiple levels of abstraction, in many object-oriented analysis methodologies such as UML and OMT. It uses hierarchical decomposition in the analysis of objects, functionality, and dynamic behavior consistently. HOOMT provides not only modeling language elements but also structured processes and guidelines for structured object-oriented analysis.