Journal of the Japanese Society for Artificial Intelligence
Online ISSN : 2435-8614
Print ISSN : 2188-2266
Volume 12 , Issue 1
Showing 1-34 articles out of 34 articles from the selected issue
Print ISSN:0912-8085 until 2013
  • Akira ICHIKAWA
    Type: Preface
    1997 Volume 12 Issue 1 Pages 1
    Published: January 01, 1997
    Released: September 29, 2020
    MAGAZINES FREE ACCESS
    Download PDF (83K)
  • [in Japanese]
    Type: Cover article
    1997 Volume 12 Issue 1 Pages 2
    Published: January 01, 1997
    Released: September 29, 2020
    MAGAZINES FREE ACCESS
    Download PDF (77K)
  • Shuji DOSHITA
    Type: Special issue
    1997 Volume 12 Issue 1 Pages 3-12
    Published: January 01, 1997
    Released: September 29, 2020
    MAGAZINES FREE ACCESS
    Download PDF (769K)
  • Yasuhisa NIIMI
    Type: Special issue
    1997 Volume 12 Issue 1 Pages 13-17
    Published: January 01, 1997
    Released: September 29, 2020
    MAGAZINES FREE ACCESS
    Download PDF (399K)
  • Hozumi TANAKA
    Type: Special issue
    1997 Volume 12 Issue 1 Pages 18-23
    Published: January 01, 1997
    Released: September 29, 2020
    MAGAZINES FREE ACCESS
    Download PDF (496K)
  • Riichiro MIZOGUCHI
    Type: Special issue
    1997 Volume 12 Issue 1 Pages 24-29
    Published: January 01, 1997
    Released: September 29, 2020
    MAGAZINES FREE ACCESS
    Download PDF (492K)
  • Katsuhiko SHIRAI
    Type: Special issue
    1997 Volume 12 Issue 1 Pages 30-35
    Published: January 01, 1997
    Released: September 29, 2020
    MAGAZINES FREE ACCESS
    Download PDF (442K)
  • Hitoshi MATSUBARA
    Type: Corner article
    1997 Volume 12 Issue 1 Pages 36-43
    Published: January 01, 1997
    Released: September 29, 2020
    MAGAZINES FREE ACCESS
    Download PDF (634K)
  • Yoshiaki SHIRAI
    Type: Corner article
    1997 Volume 12 Issue 1 Pages 44-45
    Published: January 01, 1997
    Released: September 29, 2020
    MAGAZINES FREE ACCESS
    Download PDF (152K)
  • Kenichi MORI
    Type: Corner article
    1997 Volume 12 Issue 1 Pages 46-47
    Published: January 01, 1997
    Released: September 29, 2020
    MAGAZINES FREE ACCESS
    Download PDF (120K)
  • Yuuichi KAWAGUCHI, Kiyoshi AKAMA, Eiichi MIYAMOTO
    Type: Technical paper
    1997 Volume 12 Issue 1 Pages 48-57
    Published: January 01, 1997
    Released: September 29, 2020
    MAGAZINES FREE ACCESS

    An IS-A taxonomy and a PART-OF relation are fundamental to knowledge representation. In many logic languages, computational objects have classes (or types) in an IS-A taxonomy and are composed of their components (substructures). This paper proposes a new language for representation and calculation of logical objects which have classes and substructures. In high-level knowledge representation systems based on conventional logic, computational objects are often very complicated. In order to define their meanings, we need many concepts such as higher-order predicates (e. g. "map"), predicates (e. g. inc), functions (e. g. cons), constants (e. g. 3), variables (e. g. B), classes (or types) (e. g. human) and instances (e. g. adam). This complexity makes it difficult to improve the expressive and computational power of conventional knowledge representation systems. One of the difficulties is to formalize higher-order predicates in typed logic languages, i.e., having predicate variables as arguments in atoms may produce type mismatches. To overcome this difficulty, we reconsider the basic concepts and simplify the language by integrating some basic concepts. Thus, conventional basic concepts (constants, classes, instances, predicates, higher order predicates and functions) are integrated into a single concept. Some other features of the proposed language are: (1) A class is not a simple constant but a sequence of constants, which allows us to represent complex class structures more easily. (2) Constraints of substructures of objects in a crass can be defined by specification of the classes of substructures of the objects. (3) Class objects are introduced to handle classes as logical objects, which solves the problem of higher-order predicates. (4) The computation for the language is defined based on the extended unification of above-mentioned objects. The proposed language is a simpler and more powerful computation system than conventional typed languages based on logic.

    Download PDF (763K)
  • Kenichi YOSHIDA, Hiroshi MOTODA
    Type: Technical paper
    1997 Volume 12 Issue 1 Pages 58-67
    Published: January 01, 1997
    Released: September 29, 2020
    MAGAZINES FREE ACCESS

    Inductive learning, which tries to find rules from data, has been an important area of investigation. One major research theme of this area is the data representation language that the learning methods can use. The conventional rule learning methods use an attribute-value table as a data representation language, whereas inductive logic programming (ILP) uses the first-order logic. We propose colored directed graph as a data representation language for inductive learning methods. Graph-based induction (GBI) uses this data representation language. The expressiveness of graph stands between the attribute-value table and the first-order logic. Thus its learning potential is weaker than that of ILP, but stronger than that of the conventional attribute-value learning methods. The real advantage of GBI appears in the domain where the dependency between data bears the essential information. The behavior analysis of computer users is a typical example of such a domain. In this domain, the complex structure of dependency between the user tasks prevents us from using the conventional attribute-value learning methods, and ILP cannot meet the requirement for the efficiency. In this paper, we explain GBI method and give experimental results. We also discuss the relationship between this new method and conventional inductive inference methods such as conventional classification rule learning methods, constructive induction methods, inductive logic programming methods, macro rule learning methods, and concept learning methods. While this list of the methods covers the wide area of inductive inference, we find that most of them can use Stepwise Pair Expansion as their basic algorithm. The use of the pairs and the representation language define the function and characteristics of each method. We also discuss the use of the statistical measures such as gini index and information gain index to realize various inductive inference functions.

    Download PDF (817K)
  • Toshiyuki MATSUO, Toyoaki NISHIDA, Ken'ichi HOSHIMOTO
    Type: Technical paper
    1997 Volume 12 Issue 1 Pages 68-77
    Published: January 01, 1997
    Released: September 29, 2020
    MAGAZINES FREE ACCESS

    In this paper, we describe a practical method of extracting, structuring, summarizing, and integrating technical information from technical papers in metallurgy. The heart of the method is packets of domain specific knowledge called KP (Knowledge Pieces) in which procedures for extracting and structuring technical information from technical papers are embedded. We studied information structure of ten technical papers in metallurgy and constructed about a hundred KPs. We implemented a system called METIS which takes technical papers in metallurgy encoded in a mark-up language and produces a varieties of summaries and surveys including structured technical summary, visual display of similarites and differences of relevant papers, and Cause-effect relations. We have undertaken qualitative and quantitative evaluation of METIS against 106 technical papers so far. The evaluation demonstrates the reliability and robustness of our method.

    Download PDF (772K)
  • Kazuteru MIYAZAKI, Masayuki YAMAMURA, Shigenobu KOBAYASHI
    Type: Technical paper
    1997 Volume 12 Issue 1 Pages 78-89
    Published: January 01, 1997
    Released: September 29, 2020
    MAGAZINES FREE ACCESS

    Reinforcement learning is a kind of machine learning. It aims to adapt an agent to a given environment with a clue to rewards. Profit sharing (PS) can get rewards efficiently at an initial learning phase. However, it can not always learn an optimum policy that maximizes rewards per an action. Though Q-learning is guaranteed to obtain an optimum policy, it needs numerous trials to learn it. On Markov decision processes (MDPs), if a correct environment model is identified, we can derive an optimum policy by applying Policy Iteration Algorithm (PIA). As an efficient method for identifying MDPs, k-Certainty Exploration Method has been proposed. We consider that ideal reinforcement learning systems are to get some rewards even at an initial learning phase and to get mere rewards as the identification of environments proceeds. In this paper, we propose a unified learning system : MarcoPolo which considers both getting rewards by PS or PIA and identifying the environment by k-Certainty Exploration Method. MarcoPolo can realize any tradeoff between exploitation and exploration through the whole learning process. By applying MarcoPolo to an example, its basic performance is shown. Moreover, by applying it to Sutton's maze problem and its modified version, its feasibility on more realistic domains is shown.

    Download PDF (979K)
  • Shigeo MATSUBARA, Toru ISHIDA
    Type: Technical paper
    1997 Volume 12 Issue 1 Pages 90-99
    Published: January 01, 1997
    Released: September 29, 2020
    MAGAZINES FREE ACCESS

    We propose a real-time planning algorithm called RTSS (Real-Time Search with Subgoaling), which incorporates the STRIPS subgoaling function into a real-time search algorithm called RTA*. This algorithm interleaves subgoaling and real-time search processes in the way that it continues to make subgoals until the subgoal becomes simple enough for an RTA*'s heuristic function, then applies RTA* to the subproblem. RTSS can overcome the drawbacks of RTA* and STRIPS in real-time problem solving, i. e., the algorithm does not lead to a blind search in contrast to the other two. RTSS includes the function of flexibly changing subgoal sequences in order to utilize information obtained during problem solving in dynamic environments. By an analysis using a simple model, we show that the search cost can be significantly reduced by switching between subgoaling and real-time search. Furthermore, experiments on robot task planning problems show that RTSS can attain the goal without performing many superfluous actions, while RTA* and STRIPS often tend to perform a blind search that fails to attain the goal.

    Download PDF (815K)
  • Akimoto KAMIYA, Isao ONO, Masayuki YAMAMURA, Shigenobu KOBAYASHI
    Type: Technical paper
    1997 Volume 12 Issue 1 Pages 100-110
    Published: January 01, 1997
    Released: September 29, 2020
    MAGAZINES FREE ACCESS

    Power plant start-up scheduling is aimed mainly at minimizing the start-up time of both boiler and turbine, while limiting turbine rotor stresses to acceptable values. This problem, with a number of local optima, can be formulated as a combinatorial optimization problem. In order to find the optimal or near-optimal start-up schedule efficiently, we applied evolutionary optimization techniques-Genetic Algorithms (GA)-with an "enforcement operator" and "reuse function" ( [Kamiya 95] in English). The enforcement operator is to limit the search of GA-combined with local search strategy-near the boundary of the feasible solution space, where the optimal solution is supposed to exist. The reuse function is to memorize the simulation results of the objective function of those previously generated solutions, and to reuse them whenever an identical solution is generated. In this paper, in order to increase the search efficiency further, we extend our proposed framework to integrate tabu strategy with the local search GA. The tabu strategy is to forbid some moves at a present iteration in order to avoid cycling and to make early escape from a local optimal point possible. Test results suggest that GA integrated with tabu strategy has the best performance among stand-alone GA, stand-alone tabu search and simulated annealing with or without tabu strategy. In addition, the optimized solution reduces the start-up time by approximately 10%, or 20 minutes, relative to conventional methods.

    Download PDF (914K)
  • Tomohiro NAKATANI, Masataka GOTO, Takeshi KAWABATA, Hiroshi G. OKUNO
    Type: Technical paper
    1997 Volume 12 Issue 1 Pages 111-120
    Published: January 01, 1997
    Released: September 29, 2020
    MAGAZINES FREE ACCESS

    This paper presents the Residue-Driven Architecture (RDA) as a general computational frame-work for sound stream segregation based on a multi-agent paradigm. Sound stream segregation is an important primary processing for computationally understanding sounds (Computational Auditory Scene Analysis) in the real-world. Since RDA is designed without assuming any specific sound attributes, it can be applied to various kinds of sound stream segregation problems. The RDA consists of three kinds of agents : an event-detector, a tracer-generator, and tracers. The event-detector calculates a residue by subtracting the predicted input from the actual input. When a residue exceeds a threshold value, tracer-generator generates a tracer that extracts a sound stream from the residue and returns a predicted input of the next time frame to the event-detector. The RDA is applied to the design of two subsystems : A monaural subsystem segregates sound streams under background noise using harmonic structure ; a binaural subsystem refines the sound streams segregated by the monaural system using the direction of the sound source. These subsystems can be concisely designed and simply implemented based on the RDA ; therefore, the effectiveness of the RDA is proven. In addition, experimental results show that the capability of the sound stream segregation system is improved by combining these subsystems.

    Download PDF (762K)
  • Yoko ISINO, Takao TERANO
    Type: Technical paper
    1997 Volume 12 Issue 1 Pages 121-131
    Published: January 01, 1997
    Released: September 29, 2020
    MAGAZINES FREE ACCESS

    Marketing decision making tasks require the acquisition of efficient decision rules from noisy questionnaire data. Unlike popular learning-from-example methods, in such tasks, we must interpret the characteristics of the data without clear features of the data nor predetermined evaluation criteria. The problem is how domain experts get simple, easy-to-understand, and accurate knowledge from noisy data. This paper describes a novel method to acquire efficient decision rules from questionnaire data using both simulated breeding and machine learning techniques. The basic ideas of the method are that simulated breeding is used to get the effective features from the questionnaire data and that machine learning is used to acquire simple decision rules from the data. The simulated breeding is one of the Genetic Algorithm based techniques to subjectively or interactively evaluate the qualities of offspring generated by genetic operations in a human-in-a-loop manner. The proposed method has been qualitatively and quantitatively validated by a case study on consumer product questionnaire data: the acquired rules are simpler than the results from the direct application of inductive learning; a domain expert admits that they are adequate and easy to understand; and they are at the same level on the accuracy compared with the other methods. The prerequisites of the method are so simple that it can be used to various decision making problems.

    Download PDF (895K)
  • Yoshinobu KITAMURA, Shinji YOSHIKAWA, Munehiko SASAJIMA, Mitsuru IKEDA ...
    Type: Technical paper
    1997 Volume 12 Issue 1 Pages 132-143
    Published: January 01, 1997
    Released: September 29, 2020
    MAGAZINES FREE ACCESS

    This research is concerned with causal understanding and qualitative reasoning of behavior of physical systems, which are crucial issues of model-based problem solving. In this paper, we describe a domain ontology of fluid systems and an ontology of time for generating causal ordering in terms of components. Our ontology design has been done according to the following three criteria : cognitive causal explanations in terms of components, reusability of components and disambiguation of reasoning results. We discuss in-depth requirements for domain ontologies and propose a causal specification scheme to represent component's local causal properties and an ontology of time to enable intuitive causal ordering of complex behavior originated in the combination of components. We identify causality of fluid systems following the requirements and describe reusable models of crucial components of plants and general properties of fluid and heat for deriving global knowledge. A method of qualitative reasoning and causal ordering is discussed together with its capability and mechanism. The ontologies have been successfully applied to nuclear power plant modeling and its qualitative simulation. Reasoning results matched those obtained by domain experts.

    Download PDF (1067K)
  • Hiroyoshi WATANABE, Kenzo OKUDA
    Type: Technical paper
    1997 Volume 12 Issue 1 Pages 144-151
    Published: January 01, 1997
    Released: September 29, 2020
    MAGAZINES FREE ACCESS

    One of the important issues of case-based reasoning is its adaptability for the change of environmental conditions. This means that cases which were useful for solving problems in the past often become unsuitable since environmental conditions would change as time passes. To solve such a problem, we have proposed a method of case management based on 'forgetting', which is supposed to imitate the human memory for removing obsolete cases. Though the method is very effective in adapting the case base to the environment, imitating the human memory may not always be the best method. For instance, while the system we implemented extends the term of remembrance of cases which are recalled during solving problems, a strategy which removes those cases is also plausible because a solution of a current problem is saved as a new case to the case base and it can cover for the old case which is utilized in solving the problem. In this paper, we propose a strategy for forgetting cases for a memory-restricted forgetting mechanism which is implemented by defining the size of the case base, i. e. the maximum number of cases is N, and replacing an old case with a new case when the number of cases reached N. We describe three basic strategies for the forgetting mechanism and propose extended forgetting strategies, i. e. ・ combination of the basic strategies and ・ the strategy which uses heuristics. The effectivness of the proposed strategies for improving the performance of case-based reasoning systems is demonstrated through simulations in the electric power systems.

    Download PDF (644K)
  • Keiji SUZUKI, Jun YOSHIMURA, Yukinori KAKAZU
    Type: Technical paper
    1997 Volume 12 Issue 1 Pages 152-159
    Published: January 01, 1997
    Released: September 29, 2020
    MAGAZINES FREE ACCESS

    This paper discusses the organizations of multiple agents for task processing problems and the properties of action selection mechanisms for the organizations. In particular, we introduce the multiple monkey banana problem that is extended from the original monkey banana problem. In this problem, each monkey as an agent tries to ride on other agents for taking the above bananas. Namely, the agents have to build the Organizations. It is obvious that the processes of making organizations include the many deadlock states. Thus the agents are not only to behave autonomously according to the sensing of the environment but also to have the property of persistency and abundance in the action selection mechanisms for escaping the deadlock states. To realize the properties, we introduce the fatigue parameters in the behavior network. The fatigue parameters act as short memory of sensing inputs and outputs of actions in each node. The proposed network builds as hierarchical architecture. In this architecture, we consider the two types of the strategies. One is almost opportunity strategy. Another strategy uses the organizational sensing inputs. Throughout the introduced problem, we examine the effects of the fatigue parameters and the strategies in the proposed behavior network for the organization problems.

    Download PDF (661K)
  • [in Japanese]
    Type: Other
    1997 Volume 12 Issue 1 Pages 160-161
    Published: January 01, 1997
    Released: September 29, 2020
    MAGAZINES FREE ACCESS
    Download PDF (165K)
  • [in Japanese]
    Type: Corner article
    1997 Volume 12 Issue 1 Pages 162-163
    Published: January 01, 1997
    Released: September 29, 2020
    MAGAZINES FREE ACCESS
    Download PDF (133K)
  • [in Japanese]
    Type: Corner article
    1997 Volume 12 Issue 1 Pages 164
    Published: January 01, 1997
    Released: September 29, 2020
    MAGAZINES FREE ACCESS
    Download PDF (85K)
  • [in Japanese]
    Type: Corner article
    1997 Volume 12 Issue 1 Pages 165
    Published: January 01, 1997
    Released: September 29, 2020
    MAGAZINES FREE ACCESS
    Download PDF (70K)
  • [in Japanese]
    Type: Corner article
    1997 Volume 12 Issue 1 Pages 166
    Published: January 01, 1997
    Released: September 29, 2020
    MAGAZINES FREE ACCESS
    Download PDF (96K)
  • Type: Activity report
    1997 Volume 12 Issue 1 Pages 167-170
    Published: January 01, 1997
    Released: September 29, 2020
    MAGAZINES FREE ACCESS
    Download PDF (214K)
  • Type: Activity report
    1997 Volume 12 Issue 1 Pages 171-174
    Published: January 01, 1997
    Released: September 29, 2020
    MAGAZINES FREE ACCESS
    Download PDF (252K)
  • Type: Activity report
    1997 Volume 12 Issue 1 Pages 175-176
    Published: January 01, 1997
    Released: September 29, 2020
    MAGAZINES FREE ACCESS
    Download PDF (149K)
  • Type: Activity report
    1997 Volume 12 Issue 1 Pages b001-b012
    Published: January 01, 1997
    Released: September 29, 2020
    MAGAZINES FREE ACCESS
    Download PDF (824K)
  • Type: Cover page
    1997 Volume 12 Issue 1 Pages c001
    Published: January 01, 1997
    Released: September 29, 2020
    MAGAZINES FREE ACCESS
    Download PDF (24K)
  • Type: Cover page
    1997 Volume 12 Issue 1 Pages c001_2
    Published: January 01, 1997
    Released: September 29, 2020
    MAGAZINES FREE ACCESS
    Download PDF (77K)
  • Type: Table of contents
    1997 Volume 12 Issue 1 Pages i001
    Published: January 01, 1997
    Released: September 29, 2020
    MAGAZINES FREE ACCESS
    Download PDF (60K)
  • Type: Table of contents
    1997 Volume 12 Issue 1 Pages i001_2
    Published: January 01, 1997
    Released: September 29, 2020
    MAGAZINES FREE ACCESS
    Download PDF (70K)
feedback
Top