Catálogo de publicaciones - libros

On behalf of the program committee (PC) of JSAI 2003, I would like to thank all the chair persons, discussants, and attentive audience who contributed to select these awarded papers.

A machine learning technique called Graph-Based Induction (GBI) efficiently extracts typical patterns from graph-structured data by stepwise pair expansion (pairwise chunking). It is very efficient because of its greedy search. We have expanded GBI to construct a decision tree that can handle graph-structured data. DT-GBI constructs a decision tree while simultaneously constructing attributes for classification using GBI. In DT-GBI attributes, namely substructures useful for classification task, are constructed by GBI on the fly during the tree construction. We applied both GBI and DT-GBI to classification tasks of a real world hepatitis data. Three classification problems were solved in five experiments. In the first 4 experiments, DT-GBI was applied to build decision trees to classify 1) cirrhosis and non-cirrhosis (Experiments 1 and 2), 2) type C and type B (Experiment 3), and 3) positive and negative responses of interferon therapy (Experiment 4). As the patterns extracted in these experiments are thought discriminative, in the last experiment (Experiment 5) GBI was applied to extract descriptive patterns for interferon therapy. The preliminary results of experiments, both constructed decision trees and their predictive accuracies as well as extracted patterns, are reported in this paper. Some of the patterns match domain experts’ experience and the overall results are encouraging.

In this paper, we study an online data mining problem from streams of semi-structured data such as XML data. Modeling-semistructured data and patterns as labeled ordered trees, we present an online algorithm StreamT that receives fragments of an unseen possibly infinite semi-structured data in the document order through a data stream, and can return the current set of frequent patterns immediately on request at any time. We give modifications of the algorithm to other online mining models. Furthermore we implement our algorithms in different online models and candidate management strategies, then show empirical analyses to evaluate the algorithms.

We have developed a telephone-based cooperative natural language dialogue system. Since natural language involves very various expressions, a large number of VoiceXML scripts need to be prepared to handle all possible input patterns. Thus, flexible dialogue management for various user utterances is realized by generating VoiceXML scripts dynamically. Moreover, we address the issue of appropriate user modeling to generate cooperative responses to users. Specifically, three dimensions of user models are set up: the to the system, the on the target domain and the degree of . The models are automatically derived by decision tree learning using real dialogue data collected by the system. Experimental evaluation showed that the cooperative responses adapted to individual users served as good guides for novices without increasing the duration of dialogue for skilled users.

Mobility represents very basic and essential behavior for people: reaching a destination, strolling at will, and much more. Indispensable for independence, it also makes life more enjoyable. Yet moving from place to place may be difficult for disabled, elderly, or ill individuals affected by an impairment of sight, hearing, or lower-extremity motor function, which undermine abilities needed for mobility: recognizing things, controlling motor function, and accessing information. To offset this, countries and communities have been actively preparing systems and facilities in recent years to make routes barrier-free. But it would be unfeasible to make all routes barrier-free, and there continues to be a great need for mobility support with IT technology as an alternative means of assistance. We have been researching to put Robotic Communication Terminals (RCT) into practice, which supports the three elementary behaviors of recognition, actuation, and information access, targeting almost all the pedestrians including elderly and disabled people with various types, levels, and duration of disabilities. The RCT consists of three types of terminals: “environment-embedded terminal”, “user-carried mobile terminal”, and “user-carrying mobile terminal”. These terminals communicate with one another to provide the users with a comfortable means of mobility. This paper introduces our recent research progress.

After the success of the first workshop of Agent-based approach in Economic and Social Complex Systems (AESCS) in Shimane, 2001, we have held a similar workshop in Niigata, 2003. The title is International Workshop on Agent-Based Modeling. The aims and scope are summarized as follows: Agent-Based Modeling has become one of major techniques to design and analyze complex adaptive systems including societies, economics, organizations, business management, Web applications, and the other engineering fields. The objective of the Workshop is to continue the efforts to foster the formation of an active multi-disciplinary community on multiagent, computational economics, organizational science, social dynamics, and complex adaptive systems, in conjunction with the 17th annual conference of JSAI, the largest AI related annual conference in the Pacific Asia region.

In this paper, we consider the problem of the mechanism design for the multi-agents system. We develop the social learning model for the mechanism design for creating the collective action with an efficient cost sharing rule. We consider the situation in which self-interest agents have incentives to cooperate each other for jointly acquiring the environmental level with sharing the necessary cost. We obtain the optimal level of the environment to be acquired and the cost allocation rule so that their individual rationality is satisfied, and at the same time the social rationality is also satisfied. We show that the factors such as the value (worth) of the environmental level perceived by each agent and the cost affect the level the collective action. A social rule of allocating the common cost among agents is developed with decentralized transaction mechanism. We formulate and analyze the problem of cooperating multiple agents under uncertainty. We show that when agents cooperate in order to encounter uncertainty when acting alone, their benefits would not be as attractive, and hence cooperate to share the risk. As a specific example, we consider the model of obtaining the environmental level by sharing cost. We propose the negotiation mechanism for sharing cost among agents. With that mechanism, they can learn and obtain the unbiased and fare cost distribution rule.

In this paper the Minority Game is applied to examine the effects of the common information controller for generating the cooperative behavior among the agents. If the controller modifies the information accurately, it is expected that social profit can be raised even under the situation where coopration is difficult. Through the simulation the effects and the properties of the controller are examined.

In this paper, we constructed three types of agents, which are different in efficiency and accuracy of learning. They were compared using acquired payoff in a game-theoretic situation that is called Minority game. As a result, different types of learning methods got the highest payoff according to the complexity of environmental change and learning speed.

This paper presents a new organization model that addresses the effects of networks on the sectionalism phenomenon, defined as excessive concern that members of a section have for the interests of their own section. No studies tackled the relationship between human communication networks and sectionalism. The points of our model design are: network distributed agents with a sense of values, extended random network structures, and a new index to monitor sectionalism. A homogeneous effect of communication networks and a heterogeneous effect of sectional specialization were also introduced into the model. Empirical results showed that sectionalism behavior and the performance of the proposed index were superior to conventional indices when capturing sectional structures. Finally, we showed one example of the availability of such a multi-agent network approach. Simulation results clearly illustrated the effect of cross-sectional links on sectionalism reduction by following a so-called “power law.”