Journal of the Japanese Society for Artificial Intelligence
Online ISSN : 2435-8614
Print ISSN : 2188-2266
Volume 15 , Issue 2
Showing 1-33 articles out of 33 articles from the selected issue
Print ISSN:0912-8085 until 2013
  • Hideya HAYASHI
    Type: Preface
    2000 Volume 15 Issue 2 Pages 221
    Published: March 01, 2000
    Released: September 29, 2020
    MAGAZINES FREE ACCESS
    Download PDF (90K)
  • Hitoshi KANOH
    Type: Cover article
    2000 Volume 15 Issue 2 Pages 222-225
    Published: March 01, 2000
    Released: September 29, 2020
    MAGAZINES FREE ACCESS
    Download PDF (475K)
  • Toshio YAMAZAKI
    Type: Special issue
    2000 Volume 15 Issue 2 Pages 226-233
    Published: March 01, 2000
    Released: September 29, 2020
    MAGAZINES FREE ACCESS
    Download PDF (936K)
  • Masayoshi AOKI
    Type: Special issue
    2000 Volume 15 Issue 2 Pages 234-241
    Published: March 01, 2000
    Released: September 29, 2020
    MAGAZINES FREE ACCESS
    Download PDF (975K)
  • Yasuyoshi KATO
    Type: Special issue
    2000 Volume 15 Issue 2 Pages 242-250
    Published: March 01, 2000
    Released: September 29, 2020
    MAGAZINES FREE ACCESS
    Download PDF (973K)
  • Hitoshi IBA
    Type: Corner article
    2000 Volume 15 Issue 2 Pages 251-258
    Published: March 01, 2000
    Released: September 29, 2020
    MAGAZINES FREE ACCESS
    Download PDF (838K)
  • Isao ONO, Masayuki YAMAMURA, Hajime KITA
    Type: Corner article
    2000 Volume 15 Issue 2 Pages 259-266
    Published: March 01, 2000
    Released: September 29, 2020
    MAGAZINES FREE ACCESS
    Download PDF (928K)
  • Hajime KIMURA, Shigenobu KOBAYASHI
    Type: Technical paper
    2000 Volume 15 Issue 2 Pages 267-275
    Published: March 01, 2000
    Released: September 29, 2020
    MAGAZINES FREE ACCESS

    We present an analysis of actor-critic algorithms, in which the actor updates its policy using eligibility traces of the policy parameters. Most of the theoretical results for eligibility traces have been for only critic's value iteration algorithms. This paper investigates what the actor's eligibility trace does. The results show that the algorithm is an extension of Williams' REINFORCE algorithms for infinite horizon reinforcement tasks, and then the critic provides an appropriate reinforcement baseline for the actor. Thanks to the actor's eligibility trace, the actor improves its policy by using a gradient of actual return, not by using a gradient of the estimated return in the critic. It enables the agent to learn a fairly good policy under the condition that the approximated value function in the critic is hopelessly inaccurate for conventional actor-critic algorithms. Also, if an accurate value function is estimated by the critic, the actor's learning is dramatically accelerated in our test cases. The behavior of the algorithm is demonstrated through simulations of a linear quadratic control problem and a pole balancing problem.

    Download PDF (904K)
  • Takashi ONODA, Gunnar RATSCH, Klaus R. MULLER
    Type: Technical paper
    2000 Volume 15 Issue 2 Pages 287-296
    Published: March 01, 2000
    Released: September 29, 2020
    MAGAZINES FREE ACCESS

    Recent work has shown that combining multiple versions of weak classifiers such as decision trees or neural networks results in reduced test set error. However, the analysis and the theory in reducing generalization error has not been well understood. To study this in greater detail, we analyze the asymptotic behavior of AdaBoost type algorithms. The theoretical analysis establishes the relation between the distribution of margins of the training examples and the generated voting classification rule. This paper shows asymptotic experimental results for the binary classification case underlining the theoretical findings. Finally, we point out that AdaBoost type algorithms lead to overfitting and we improve AdaBoost by in-troducing regularization to avoid overfitting and to thereby reduce the generalization error. Also we show in numerical experiments that our improvement can lead to supperior classification results.

    Download PDF (1072K)
  • Atsushi UENO, Shinichi NAKASUKA, Koichi HORI
    Type: Technical paper
    2000 Volume 15 Issue 2 Pages 297-308
    Published: March 01, 2000
    Released: September 29, 2020
    MAGAZINES FREE ACCESS

    Real robots should be able to adapt autonomously to various environments in order to go on executing their task without a break. For this purpose, they should be able to learn how to abstract useful information from a huge amount of information in the environment while executing their task. This paper proposes a new architecture which performs categorical learning and behavioral learning parallelly with task execution. We call the architecture Situation Transition Network System (STNS). In categorical learning, it makes a flexible state representation and modifies it according to the results of behaviors. Behavioral learning is reinforcement learning on the state representation. Simulation results have shown that this architecture can learn efficiently and adapt to unexpected changes of the environment autonomously.

    Download PDF (1417K)
  • Yutaka IMAI, Shun ISHIZAKI
    Type: Technical paper
    2000 Volume 15 Issue 2 Pages 309-320
    Published: March 01, 2000
    Released: September 29, 2020
    MAGAZINES FREE ACCESS

    In this paper, we propose a new automatic extraction method to obtain salient features from a metaphorical expression consisting of two noun concepts as target T and source S concepts, "T is S". Human cognitive experiments were carried out to determine the initial feature values in the calculation procedure. First, we obtained a bundle of attributes from human association experiments on concepts. Second, using the bundle, we implemented SD(Semantic Differential ) Technique experiments. We selected 36 adjective pairs such as "wide-narrow", "clean-dirty", "strong-weak", and so on.The salient feature extraction method is carried out in the following procedure. First, using rated results from the SD Technique experiments, the features of the target concept bundle are sorted and allocated to corresponding axes in a circular shaped diagram. Each axis corresponds to one of the absolute average values of the rated values of features. Next, corresponding to the sorted target concept bundle, the features of the source bundle are allocated in the circular shaped diagram. Finally, the salient features are extracted as vertices of the convex hull with respect to a set of points indicating the combined features of the target and source concepts. the calculation procedure for the automatic extraction of the salient features is described by examples. Since the circular shaped diagram is used for visualization, the extracted salient features is understandable easily and intuitively.

    Download PDF (1188K)
  • Takashi OKADA
    Type: Technical paper
    2000 Volume 15 Issue 2 Pages 321-330
    Published: March 01, 2000
    Released: September 29, 2020
    MAGAZINES FREE ACCESS

    The cascade model is proposed as a new method for finding discrimination rules. The method is based on the itemset lattice used in association rule mining. Items arising from explanation attributes are employed to construct the lattice, while the distribution of the class values of the supporting instances is attached to each node in the lattice. The gini-index is used to indicate the potential of the node. The introduction of node potential suggests an image of cascades, with nodes as lakes and links as waterfalls, in which an instance corresponds to a drop of water falling down the cascades. The power of a link is analogous to the hydroelectric power of a waterfall. The problem of finding discrimination rules is then formulated as a search for powerful waterfalls in the cascades. the model has been implemented as the DISCAS system, using a new pruning criterion to avoid combinatorial explosion of the number of nodes in the lattice. Algorithms for node generation and rule extraction are described. Application to House voting records shows that the resulting rules are simple and comprehensible.

    Download PDF (966K)
  • Fumiaki IMANARI, Manabu YAMAUCHI, Hitoshi OGAWA
    Type: Technical paper
    2000 Volume 15 Issue 2 Pages 331-338
    Published: March 01, 2000
    Released: September 29, 2020
    MAGAZINES FREE ACCESS

    Since former studies about schedule problems have focused on suitability and efficiency of schedules under static environment, they can not be adapted to dynamic environment under which unpredictable delay of operations occurs.In order to cope with this problem, we propose a new robust scheduling method with "Acceptable Delay Time(ADT)". An acceptable delay time of an operation means the time within which if the time of delay of the operation is, all jobs finish until their due dates. We introduce ADT in order to finish as many jobs within their due dates as possible when delay of operations occurs. We try to generate robuster schedules by adapting the scheduling method based on the order constraint and iteratively improving an initial schedule with an evaluation function using ADT.With the experiment of the job shop scheduling problem with delay of operations, we examine robustness of the schedule improved with three different kinds of evaluation functions using ADT and compare them with one using the length of the schedule. As a result of the experiment, we show that as schedules have the longer average of ADT, more unfinished jobs within their due dates decrease.

    Download PDF (784K)
  • Yasuo KUDO, Tetsuya MURAI, Tsutomu DA-TE
    Type: Technical paper
    2000 Volume 15 Issue 2 Pages 339-347
    Published: March 01, 2000
    Released: September 29, 2020
    MAGAZINES FREE ACCESS

    We propose modal logics of ordered arrows to represent belief update as logical calculus of modal logic. Belief update, proposed by Katsuno and Mendelzon, is a theoretical formulation of the way we change our beliefs by getting or losing information based on changes of the world. Modal logics of ordered arrows are the extension of modal logics of arrows proposed by Vakarelov. We propose ordered arrow frames (OA frames, for short) that represent structures of state transition diagrams and relative plausibility of state transitions. OA frames provide Kripke semantics for modal logics of ordered arrows (OA models, for short). We also propose an axiomization of the logic of all OA frames and call this logic OAL(ordered arrow logic). OAL contains all axioms and inference rules of BAL(basic arrow logics). We prove that OAL is sound and complete for the class of all OA models. Moreover, we represent belief update operations and postulates as formulae in OAL, which means that belief update is characterized as logical calculus of OAL.

    Download PDF (1010K)
  • Makoto YOKOO, Katsutoshi HIRAYAMA
    Type: Technical paper
    2000 Volume 15 Issue 2 Pages 348-354
    Published: March 01, 2000
    Released: September 29, 2020
    MAGAZINES FREE ACCESS

    A distributed constraint satisfaction problem can formalize various application problems in Multiagent systems, and several algorithms for solving this problem have been developed. One limitation of these algorithms is that they assume each agent has only one local variable. Although simple modifications enable these algorithms to handle multiple local variables, obtained algorithms are neither efficient nor scalable to larger problems.We develop a new algorithm that can handle multiple local variables efficiently, which is based on the asynchronous weak-commitment search algorithm. In this algorithm, a bad local solution can be modified without forcing other agents to exhaustively search local problems. Also, the number of interactions among agents can be decreased since agents communicate only when they find local solutions that satisfy all of the local constraints. Experimental evaluations show that this algorithm is far more efficient than an algorithm that uses the prioritization among agents.

    Download PDF (753K)
  • Katsutoshi HIRAYAMA, Makoto YOKOO
    Type: Technical paper
    2000 Volume 15 Issue 2 Pages 355-361
    Published: March 01, 2000
    Released: September 29, 2020
    MAGAZINES FREE ACCESS

    We present resolvent-based learning as a new nogood learning method for a distributed constraint satisfaction algorithm. This method is based on a look-back technique in the CSP literature and can efficiently make effective nogoods.We combine the method with the asynchronous weak-commitment search algorithm (AWC) and evaluate the performance of the resultant algorithm on distributed 3-coloring problems and distributed 3SAT problems. As a result, we found that the resolvent-based learning works well compared to previous learning methods for distributed constraint satisfaction algorithms. We also found that the AWC with the resolvent-based learning is able to find a solution with fewer cycles than the distributed breakout algorithm, which was known to be the most efficient algorithm (in terms of cycles) for solving distributed CSPs.

    Download PDF (785K)
  • Ryoji ITOH, Tatsuhiro KONISHI, Yukihiro ITOH
    Type: Technical paper
    2000 Volume 15 Issue 2 Pages 362-375
    Published: March 01, 2000
    Released: September 29, 2020
    MAGAZINES FREE ACCESS

    In this paper, we propose an education system that helps programming learners by providing domainoriented-explanations that are explanations of functions of programs by vocabularies of domain worlds. We think novice programming learners feel difficulties when they can not understand whether bugs of their own programs are caused by failure in constructing algorithm or failure in encoding it. Therefore our system explains domain-oriented-functions of programs independently of detailed logic of the programs. By comparing the explanations with supposed behaviors of the programs, learners can check algorithms and find algorithmic bugs without middling with detailed logic of the programs. The system also shows sets of statements that realize each domain-oriented-function in order to help learners to find buggy codes. Moreover it can help learners to understand model programs by explaining them. In this paper, we describe a method of program analysis based on simulation. Our system has a domain model which represents states of a domain world and simulates a program on the model to extract its domain-oriented-functions. We also describe a method to make explanations of extracted functions and logic of programs. We have implemented an experimental system and a domain model for exercises that pays attention to numerical order such as 'greater' or 'lesser' (ex. exercise of sorting). We show our experimental system and experimental results.

    Download PDF (1648K)
  • [in Japanese]
    Type: Corner article
    2000 Volume 15 Issue 2 Pages 376-377
    Published: March 01, 2000
    Released: September 29, 2020
    MAGAZINES FREE ACCESS
    Download PDF (154K)
  • [in Japanese], [in Japanese]
    Type: Corner article
    2000 Volume 15 Issue 2 Pages 378-379
    Published: March 01, 2000
    Released: September 29, 2020
    MAGAZINES FREE ACCESS
    Download PDF (237K)
  • [in Japanese]
    Type: Corner article
    2000 Volume 15 Issue 2 Pages 380
    Published: March 01, 2000
    Released: September 29, 2020
    MAGAZINES FREE ACCESS
    Download PDF (117K)
  • [in Japanese]
    Type: Corner article
    2000 Volume 15 Issue 2 Pages 381
    Published: March 01, 2000
    Released: September 29, 2020
    MAGAZINES FREE ACCESS
    Download PDF (158K)
  • [in Japanese]
    Type: Corner article
    2000 Volume 15 Issue 2 Pages 382
    Published: March 01, 2000
    Released: September 29, 2020
    MAGAZINES FREE ACCESS
    Download PDF (154K)
  • [in Japanese]
    Type: Corner article
    2000 Volume 15 Issue 2 Pages 383
    Published: March 01, 2000
    Released: September 29, 2020
    MAGAZINES FREE ACCESS
    Download PDF (141K)
  • [in Japanese]
    Type: Corner article
    2000 Volume 15 Issue 2 Pages 384-385
    Published: March 01, 2000
    Released: September 29, 2020
    MAGAZINES FREE ACCESS
    Download PDF (211K)
  • [in Japanese]
    Type: Other
    2000 Volume 15 Issue 2 Pages 386
    Published: March 01, 2000
    Released: September 29, 2020
    MAGAZINES FREE ACCESS
    Download PDF (135K)
  • Type: Activity report
    2000 Volume 15 Issue 2 Pages 387-390
    Published: March 01, 2000
    Released: September 29, 2020
    MAGAZINES FREE ACCESS
    Download PDF (323K)
  • Type: Activity report
    2000 Volume 15 Issue 2 Pages 391-393
    Published: March 01, 2000
    Released: September 29, 2020
    MAGAZINES FREE ACCESS
    Download PDF (248K)
  • Type: Activity report
    2000 Volume 15 Issue 2 Pages b001-b016
    Published: March 01, 2000
    Released: September 29, 2020
    MAGAZINES FREE ACCESS
    Download PDF (1264K)
  • Type: Cover page
    2000 Volume 15 Issue 2 Pages c002
    Published: March 01, 2000
    Released: September 29, 2020
    MAGAZINES FREE ACCESS
    Download PDF (157K)
  • Type: Cover page
    2000 Volume 15 Issue 2 Pages c002_2
    Published: March 01, 2000
    Released: September 29, 2020
    MAGAZINES FREE ACCESS
    Download PDF (317K)
  • Type: Table of contents
    2000 Volume 15 Issue 2 Pages i002
    Published: March 01, 2000
    Released: September 29, 2020
    MAGAZINES FREE ACCESS
    Download PDF (72K)
  • Type: Table of contents
    2000 Volume 15 Issue 2 Pages i002_2
    Published: March 01, 2000
    Released: September 29, 2020
    MAGAZINES FREE ACCESS
    Download PDF (70K)
  • Type: Other
    2000 Volume 15 Issue 2 Pages o001
    Published: March 01, 2000
    Released: September 29, 2020
    MAGAZINES FREE ACCESS
    Download PDF (51K)
feedback
Top