Volume3, Number1, 2002


 

 

Title:

Fault-Diagnosis for a Class of Distributed Control Multistage Interconnection Networks

Author:

Yanggon Kim

Abstract:

    A class of (2log2N - 1)-stage interconnection networks is suitable for fault-tolerance with effective use of the redundant paths which provide multiple disjoint paths between each pair of input/output terminals. This paper focuses on a new fault-diagnosis method for a class of distributed (2log2N - 1)-stage interconnection networks. To perform this fault-diagnosis method, the self-routing tags based on a given multistage interconnection network are generated for each phase. Six tests with self-routing tags can detect a faulty switching element and a link stuck fault in a (2log2N - 1)-stage interconnection network. Regarding the fault location problem, the fault location procedure to use O(log2log2N) tests for a distributed control network is proposed.

 

 

Title:

A Group Buy Protocol based on Coalition Formation for Agent-mediated E-Commerce

Author:

Takayuki Ito, Hiroyuki Ochi and Toramatsu Shintani

Abstract:

    Electronic commerce is becoming an increasingly im-portant channel for retail commerce. There are a large number of e-commerce sites on the Internet. Intelligent agents have been studied very widely in the field of arti-ficial intelligence and multi-agent systems. There have been several researches on agent-mediated electronic commerce. In particular, we focus on volume discount mechanism based on agent-mediated electronic com-merce. In this paper, we propose a new group buy scheme for agent-mediated electronic markets. In this scheme, we introduce the concept of sequential decision-making for seller agents and buyer agents. Further, we propose a cooperation mechanism among seller agents based on exchange items in stock. We implemented an agent-mediated electronic markets based on the group buy scheme. The result shows that the new group buy scheme is effective in supporting electronic commerce based on volume discount.

 

 

Title:

Design the Agent Model for Multi-Threaded Processes

Author:

Kazunori Iwata , Nobuhiro Ito and Naohiro Ishii

Abstract:

    The purpose of this paper is to model the agents which can collaborate in the dynamically-changing environment. In the environment, the agent needs a number of functionality to describe a dynamically planning agent and to organize groups with other agents [1-7].

    In order to model the agent, which can accomplish these characteristics, we have focused on -calculus [8,9]. -calculus is a process calculus to describe a channel-based communication among distributed processes and provides dynamically changing structures and concurrent processes. In -calculus processes have a property to choose one process from concurrent processes. The property provides flexibility to agents [10-14]. In order to choose one process, processes get a mutex-lock and execute the process. However, the mutex-lock is very complex to avoid deadlock [15]. Hence, we adjust the situations on the communication among the processes and have designed a -model. Moreover, we have defined the protocol and have explained how to avoid the deadlock

 

 

Title:

Self-Calibrated Neural Network Based Photometric Stereo

Author:

Yuji Iwahori, Yumi Watanabe, Kenji Funahashi and Robert J. Woodham

Abstract:

    This paper proposes a new approach of neural network based photometric stereo for a rotational object with non-uniform reflectance factor. For a rotational object, four input images are acquired under different conditions of illumination. Here, we use no calibration object to recover the shape of a target object. This is achieved by the self-calibration using the geometrical constraint from the rotation of object. The radial basis function (RBF) neural network is used to do non-parametric functional approximation. The training data to the neural network is obtained from the target object itself. The entire approach is quite empirical and no explicit assumptions are used for the surface reflectance or the directions of light source. Experiments on real data are described.

 

 

Title:

A Scheme on Software Process Component Reuse using Product-Line Practice

Author:

Jooyoung Seo, Yoonjung Lee and Byoungju Choi

Abstract:

    This paper applies the concept of product line to systematically embody the application specific software process by 1) proposing a “scheme to reuse software process”, 2) implementing “software process generating tool” for its automation.

     Based on the concept of product line, our “scheme to reuse software process” analyzes the standards, the commonalities and differences between methodologies and development processes of domains to design core assets into CBD-oriented components, and hence enables these core assets easily to create application specific software process utilizing customization pattern of components.

    Our "software process generating tool” stores the core assets developed during the core asset development phase for the purpose of reuse in storage, and forms an application specific process by automating the product development phase.

 

 

Title:

Adapting to Subsequent Changes of Environment by Learning Policy Preconditions

Author:

Tohgoroh Matsui, Nobuhiro Inuzuka and Hirohisa Seki

Abstract:

    This paper describes a method which senses a changing environment by collecting failed instances, uses concept learning for acquiring a precondition for a control policy, and partially modifies the policy for adapting to subsequent changes of the environment by reinforcement learning. A precondition for a policy represents a condition to reach a goal using the policy. Our method learns a precondition for a policy from instances of the policy's success or failure by concept learning. Using concept learning, our method has the ability to improve its behavior in states not experienced by the robot. We experimented using our method on a profit-sharing reinforcement learning system, and a decision tree learning system, C4.5. It adapted to a changing environment faster than the re-learning methods. In addition, we have confirmed that concept learning provides a method that adapts effectively to a changing environment.

 

 

Title:

Exploiting Both Spatial And Temporal Locality In Page Replacement Algorithms

Author:

Xianghui Liu ,Nohpill Park, Tao Li and Roger Y. Lee

Abstract:

The Least Recently Used (LRU) page replacement algorithm is a good approximation to the optimal replacement algorithm, but it is based on only the time of last reference and therefore does not take full advantage of locality behavior of program reference strings. This paper presents an extended LRU page replacement algorithm (ELRU) in an attempt to exploit spatial locality as well as temporal locality in memory reference during the program run time by adding two adaptive parameters: locality window size and locality function into the conventional LRU algorithm. When a page fault occurs, the selection of a victim page depends not only on the last reference time of the page in memory but also on the page involved in the locality window specified by the locality function. It is formally proven that ELRU is a stack algorithm. Extensive simulation results show that in terms of the number of page faults ELRU algorithm performs better than LRU under pure loops model and is comparable to it under independent reference model. ELRU caused some extra overhead in time and space due to the implementation of adaptation of locality window size and locality function. However, this overhead was negligible on the average.

 

 

Title:

Task Duplication and Insertion for Scheduling with Communication Costs

Author:

Nobuhiro Doi, Kazunari Sumiyoshi and Naohiro Ishii

Abstract:

An algorithm named DL(Dynamic Leveling) has been proposed by Sih and Lee [1] for scheduling precedence graphs on the distributed memory multiprocessor systems. This algorithm is one of priority-based algorithm with communication costs. In this paper, we notice how to calculate the priority of tasks. With DL, the priority is based on the earliest time the task starts. But this time is changed by task duplication, which is used to reduce the communi-cation costs. So we calculate this time on the assumption that duplications have been done and propose a refinement of task insertion to reduce processor idle times.

Experimental study shows that the quality of schedules obtained by incorporating them into scheduling algorithms, is improved shorter in the schedule length than the conventional method.

 

 

Title:

Using UML Class Diagram for RAISE Applicative Specification

Author:

N. Debnath, D. Riesco, G. Montejano, A. Dasso, A. Funes and R. Uzal

Abstract:

The objective of this research is to use the Unified Modeling Language (UML) for domain specification and to complete the semantic with a rigorous method. In this way, all the potential facilities of UML including its graphical notation, massive number of users and the potential facilities of formal methods such as RAISE can be used.

It appears that graphical notation is very useful for system specification, and recently, more and more developers have worked with UML. In this paper, we employed a graphical notation for formal specification, particularly the RAISE specification.

We present a part of UML Class Diagrams and show the meaning of a diagram with RAISE specification, especially Abstract Applicative Specification.

Then, the engineers who use a tool, such as UML/RAISE, can continue using the RAISE tools, including the RAISE proof for completing the whole specification. So a developer can use UML for the general structure of the system domain, then complete the specification using RAISE and use all verification modelling of RAISE.