人工知能学会論文誌
検索
OR
閲覧
検索
16 巻 , 6 号
選択された号の論文の11件中1~11を表示しています
    • |<
    • <
    • 1
    • >
    • >|
論文
  • 田中 卓史
    原稿種別: 研究論文
    専門分野: 基礎・理論
    16 巻 (2001) 6 号 p. 454-464
    公開日: 2002/02/28
    ジャーナル フリー
    First, we develop a type of logic grammar called Extended DCSG. This logic grammar not only defines the syntactic structures of free-word-order languages but also defines relationships between syntactic structures and their meanings. Next, we view electronic circuits as a free-word-order language, and write an Extended DCSG grammar for this language. Relationships between circuit structures and their functions are defined as grammar rules. The Extended DCSG grammar rules are converted into a logic program that parses electronic circuits and derives their functions and behaviors as meanings. These functions and behaviors are finally translated into natural language for explanation.
    抄録全体を表示
技術資料
  • 越野 亮, 岡峰 正, 木村 春彦, 広瀬 貞樹
    原稿種別: 研究論文
    専門分野: 基礎・理論
    16 巻 (2001) 6 号 p. 465-472
    公開日: 2002/02/28
    ジャーナル フリー
    Cost-based abduction, which is a technique for identifying the best explanation for a given observation based on the assumption of a set of hypothesis, is a useful knowledge processing framework for practical problems such as diagnosis, design and planning. However, the speed of reasoning of this approach is often slow. To overcome this problem, Kato et al. previously presented a more efficient cost-based abduction system, that utilized the A* search technique, however, the time and space complexities in this technique are exponential, so the identification of the optimal solution is difficult in practical applications. In this paper, we present three new systems in which the user can define the computational complexity (polynomial order), for identification of a near-optimal solution. First, we introduce two search control techniques; a real-time A* search approach in which the user can define the look-ahead depth or space, and the multi-agent real-time A* approach in which the user can define the number of real-time A* agents used in the search. We describe the implementation of three cost-based abduction reasoning systems for predicate logic knowledge bases and test the proposed systems using a diagnostic logic circuit problem. The results show that proposed systems can identify a near-optimal solution according to the predefined polynomial order of complexity, including the selection of either linear or exponential computational complexity. It is also shown that inference time and success rate are dependent on the user-defined parameters, that the three proposed systems exhibit similar performance characteristics, and that they all offer significant speed advantages over the previously described technique.
    抄録全体を表示
論文
  • 松原繁夫
    原稿種別: 研究論文
    専門分野: 基礎・理論
    16 巻 (2001) 6 号 p. 473-482
    公開日: 2002/02/28
    ジャーナル フリー
    An ascending-bid auction protocol with a fixed end time has been widely used in many Internet auction sites. In such auctions, we can observe bidders’ behavior called last minute bidding, namely, a large fraction of bids are submitted in the closing seconds of the auction. This may cause a problem of information revelation failure as well as a problem of the server’s overload and network congestion. If almost all bids are submitted only during the last minute, each bidder cannot obtain information about the good through others’ bidding behavior, which will spoil the advantage of open-bid auctions. This results in having an inefficient allocation of the good. To solve this problem, we propose a new protocol that gives bidders an incentive to fix the maximum bid of a proxy agent. We examine the property of the protocol by using game theory and clarify which situations our protocol outperforms the existing protocol by using a computer simulation.
    抄録全体を表示
  • 高橋 勇, 小西 達裕, 伊東 幸宏
    原稿種別: 研究論文
    専門分野: 基礎・理論
    16 巻 (2001) 6 号 p. 483-492
    公開日: 2002/02/28
    ジャーナル フリー
    We think that a Micro-World should have learner model in order to guide a learner who is in impasse. In other words, the system should infer which knowledge is not understood by learner (we call such knowledge ‘doubtful knowledge’). In our previous works, we proposed a planning and plan recognition approach to generate advice appropriately by observing learner’s actions in Micro-World for high school chemistry. The system can infer doubtfulness of only a type of knowledge on the relation between goal and it’s means (we call such knowledge ‘goal-means knowledge’). However, it is not enough to guide a learner appropriately. We propose the method to infer doubtfulness of the other types of knowledge (like causal relation, numerical relation and so on), in order to construct a useful learner model for helping the learner in impasse. We take two approach to infer the doubtfulness of knowledge. One is to show a learner a final goal with condition which the learner must achieve, because when the learner can keep the condition, we can think that the learner must have the knowledge necessary to keep the condition. Another is comparing learner’s faulty actions with correct actions, because difference of them are caused by learner’s lack of knowledge. In addition, we report on our experimental system which is a Micro-World for high school chemistry.
    抄録全体を表示
  • 佐藤 眞木彦, 安達 統衛
    原稿種別: 研究論文
    専門分野: 基礎・理論
    16 巻 (2001) 6 号 p. 493-500
    公開日: 2002/02/28
    ジャーナル フリー
    Schedule planning is one of the most crucial issues for any airline company, because the profit of the company directly depends on the efficiency of the schedule. This paper presents a novel scheduling method which solves problems related to time scheduling, fleet assignment and maintenance routing simultaneously by Genetic Algorithms. Every schedule constraint is embeded in the fitness function, which is described as an object oriented model and works as a simulater developing itself over time, and whose solution is executable without human correction. The schedular is able to solve the problems involving several hundred flights in a few hours, and the solutions are superior or equivalent to those by human experts in terms of the estimated profit.
    抄録全体を表示
  • 五十嵐 治一
    原稿種別: 研究論文
    専門分野: 基礎・理論
    16 巻 (2001) 6 号 p. 501-509
    公開日: 2002/02/28
    ジャーナル フリー
    In a previous paper, we proposed a solution to navigation of a mobile robot. In our approach, we formulated the following two problems at each time step as discrete optimization problems: 1) estimation of position and direction of a robot, and 2)action decision. While the results of our simulation showed the effectiveness of our approach, the values of weights in the objective functions were given by a heuristic method. This paper presents a theoretical method using reinforcement learning for adjusting the weight parameters in the objective function that includes pieces of heuristic knowledge on the action decision. In our reinforcement learning, the expectation of a reward given to a robot’s trajectory is defined as the value function to maximize. The robot’s trajectories are generated stochastically because we used a probabilistic policy for determining actions of a robot to search for the global optimal trajectory. However, this decision process is not a Markov decision process because the objective function includes an action at the previous time. Thus, Q-learning, which is a conventional method of reinforcement learning, cannot be applied to this problem. In this paper, we applied Williams’s episodic REINFORCE approach to the action decision and derived a learning rule for the weight parameters of the objective function. Moreover, we applied the stochastic hill-climbing method to maximizing the value function to reduce computation time. The learning rule was verified by our experiment.
    抄録全体を表示
  • 伊藤 一之, 松野 文俊
    原稿種別: 研究論文
    専門分野: 基礎・理論
    16 巻 (2001) 6 号 p. 510-520
    公開日: 2002/02/28
    ジャーナル フリー
    Reinforcement learning has recently received much attention as a learning method for complicated systems, e.g., robot systems. It does not need prior knowledge and has higher capability of reactive and adaptive behaviors. However increase in dimensionality of the action-state space makes it diffcult to accomplish learning. The applicability of the existing reinforcement learning algorithms are effective for simple tasks with relatively small action-state space. In this paper, we propose a new reinforcement learning algorithm: “Q-learning with Dynamic Structuring of Exploration Space Based on Genetic Algorithm ”. The algorithm is applicable to systems with high dimensional action and interior state spaces, for example a robot with many redundant degrees of freedom. To demonstrate the effectiveness of the proposed algorithm simulations of obstacle avoidance by a 50 links manipulator have been carried out. It is shown that effective behavior can be learned by using the proposed algorithm.
    抄録全体を表示
  • 尾崎 知伸, 古川 康一
    原稿種別: 研究論文
    専門分野: 基礎・理論
    16 巻 (2001) 6 号 p. 521-530
    公開日: 2002/02/28
    ジャーナル フリー
    Inductive Logic Programming (ILP) employs expressive representation languages, i.e. Prolog Programs, so that ILP can handle structural data which other traditional inductive learner can not or hardly do. In this reason, ILP has been regarded as one of the most important technologies in the area of Data Mining and Knowledge Discovery in Databases recently. However, ILP usually needs enormous computational time to obtain the results from a huge amount of data appearing in such problems as Data Mining. To cope with this problem and to make ILP more practical, we need more efficient algorithms. Since learning by ILP can be regarded as the search problem, we need to consider the reduction of the number of hypotheses to be evaluated and the efficient evaluation algorithm in order to make ILP more efficient. One of the most effective ways in reducing the hypothesis space to be searched is to compute a Most Specific Hypothesis (MSH) from one positive example by Inverse Entailment. Since MSH bounds the hypothesis space, it is important to decide which example should be used to generate MSH. In this paper, in order to reduce the computational time, we propose the following three algorithms for ILP systems which employ coverset algorithm, inverse entailment and top-down search strategy: (1)selection of search space, (2)incremental search within one class, (3)integration of hypothesis spaces among the different classes. The main common feature of these three algorithms is that, instead of single MSH, plural MSHs are considered in the search process. Experiments were conducted to assess the effectiveness of the proposed algorithms. The results show that the proposed algorithms are useful for reducing the number of candidate hypotheses to be evaluated as well as the total computational time for induction.
    抄録全体を表示
  • 尾暮 拓也, 花房 英光, 古田 一雄
    原稿種別: 研究論文
    専門分野: 基礎・理論
    16 巻 (2001) 6 号 p. 531-538
    公開日: 2002/02/28
    ジャーナル フリー
    Generally, maintenance tasks need various knowledge, which includes empirical knowledge of experienced experts. This led our interest to inheritance of maintenance expertise. We thought any powerful methodology for the purpose must be discussed, and it will never be achieved without dealing with background knowledge of maintenance. From this viewpoint, we analyzed educational documents about maintenance in order to derive representation of such background knowledge. This discussion lead to “maintenance ontology”, and a particular representation constructed with three knowledges; knowledge about design process, knowledge about trouble occurrence, and free style document. The free style document is prepared as additional description of the other two knowledges, and besides description of empirical knowledge. The empirical knowledge would be described in context of design and trouble occurrence. We have developed a prototype CAI system and asked maintenance experts to evaluate it. The result of the evaluation was totally satisfactory, and so we concluded the proposed methodology significant.
    抄録全体を表示
  • 橋本 剛, 作田 誠, 飯田 弘之
    原稿種別: 研究論文
    専門分野: 基礎・理論
    16 巻 (2001) 6 号 p. 539-547
    公開日: 2002/02/28
    ジャーナル フリー
    Brinkmate (hisshi in Japanese) is an important notion of accessing to the opponent’s King in shogi. This is essentially the same as conventional chess mating problems, where all moves are considered. However, in shogi the problem is much more difficult, as the possibilities for delivering check or threatmate, and the number of defenses are much greater.The defending side may have 200 or 300 possible defensive moves to consider. Brinkmate search is resolved into an AND/OR-tree search based on the concept of threat sequence proposedby Iida. The cost of brinkmate search is by far more expensive than mating (tsumeshogi) to find a solution. Since not much is known about brinkmate itself or brinkmate search, we first explain it by giving the definition of brinkmate. In an AND/OR tree of brinkmate search, to determine effective branches (i.e., legal moves) at any internal node is often a time-consuming task. This paper proposes a new search algorithm, denoted by SPH, for an AND/OR-tree search. The algorithm is implemented in a shogi-hisshi (Japanese-chess brinkmate-problem) program, and evaluated by testing it on difficult hisshi problems. Moreover, it is enhanced by several methods including a new idea, denoted by TDSS. The experimental results are compared with those of other programs. The program with TDSS shows the best results for solving short-step problems, while the SPH program in general outperforms the other programs in solving long-step problems.
    抄録全体を表示
  • Mofizur Rahman Chowdhury, Masayuki Numao
    原稿種別: scientific monograph
    専門分野: Foundations of AI
    16 巻 (2001) 6 号 p. 548-556
    公開日: 2002/02/28
    ジャーナル フリー
    We propose a new approach to first order inductive learning using techniques borrowed from the state of the art constructive inductive ILP systems. In this respect a learning system ALPS is presented which performs a top-down iterative broadening search through the hypothesis space. ALPS uses argument selection heuristic of constructive inductive ILP systems which enables it to avoid a huge search space. It employs an automated bias adjustment procedure through a sequence of hypothesis subspaces arranged in a hierarchical lattice. Some experiments show that in benchmark logic program synthesis tasks, ALPS visits much less search space than well-known existing algorithms which perform a hill-climbing search through the hypothesis space. ALPS is also shown to be more successful in learning situations where there exists many irrelevant background predicates and where the training set comes from an unbiased source.
    抄録全体を表示
    • |<
    • <
    • 1
    • >
    • >|
feedback
Top