Takaaki Okumoto

A cooperative search scheme for dynamic problems

The importance of building a general framework for distributed problem solving is coming to be acknowledged. Distributed search is one of such frameworks and defined as finding a required path in a given graph by cooperation of multiple agents, each of which is able to search the graph partially. In this paper, the authors propose a new cooperative search scheme for dynamic problems where costs of links are changeable in the course of the search. To cope with the dynamic character, agents cooperate with each other by exchanging cost information that they keep. When a large amount of cost information is exchanged, it improves the quality of solution, but on the other hand it raises communication overhead. It is hence significant to know how much cost information optimizes the total performance. The authors developed a testbed that simulates a communication network and applied their scheme to a routing problem which can be viewed as a dynamic problem where cost of link is defined as changeable communication delay. The authors measured its performance according to the amount of the cost information exchanged among agents.<<ETX>>

Parallel genetic algorithm on a fractal geometry-based interconnection network and its application to a classifier machine

The authors propose a parallel genetic algorithm on a multiprocessor system (FIN-1) which has a self-similarity network, and as its application, they constructed a classifier system such that the samples were classified into several classes based on the feature belonging to each sample. In the process of designing the classifier system the parallel genetic algorithm was applied to the travelling salesman problem and the sample set was classified in the Euclidean space into several categories with a measure of the distance.<<ETX>>

Conceptual clustering with parallel genetic algorithm on a multiprocessor system of a fractal geometry-based interconnection network

The methodology that classify a set of objects obtained by the observation into the meaningful groups has been proposed. In this paper, we consider the conceptual clustering using the genetic algorithm (GA) that is regarded as an effectual method to solve the combinatorial optimization problem through a multiprocessor system (FIN) which has a self-similarity network. The simulation has been implemented by the proposed method.<<ETX>>

Diffusing Inference: An Inference Method for Distributed Problem Solving

This paper presents the diffusing inference method, which is an inference method for distributed problem solving, and discuss its properties. Diffusing inference differs from other methods such as task -sharing and result-sharing, and is appropriate for applications that inherently require distributed search techniques. To discuss the algorithm and its properties rigorously, a formulation of distributed problem solving based on the state space graph is introduced. Local knowledge of each agent is represented as a partial graph in this formulation. Diffusing inference works as follows: an agent that received an initial state begins to search using only local knowledge for the goal state as far as it can. If it cannot achieve the goal, then it allocates the rest of the search to the other available agents and they continue to search likewise. Inference diffuses, as a result, among agents as distributed search proceeds. Furthermore, the property of completeness, i.e. the diffusing inference algorithm finds a solution and terminates whenever it is given a problem with a solution, is discussed and proved. Generally speaking, the performance of a diffusing inference system depends on the relation between its inference speed and its communication speed. Therefore a performance evaluation from this viewpoint is presented by using simulation techniques. It shows a guideline where the performance of diffusing inference by multiple agents is superior to that of a centralized search by a single agent.

Constrained clustering and parallel genetic algorithm on a multiprocessor system FIN

Genetic algorithms (GA) are typically regarded as the unconstrained search procedure within the given representation space. But many actual problems hold one or more constraints that must be satisfied. In this paper, we consider the incorporation of constraints into fitness function and solve the constrained clustering problem using the GA through a multiprocessor system (FIN) which has a self-similarity network.<<ETX>>