人工知能学会論文誌
検索
OR
閲覧
検索
22 巻 , 3 号
選択された号の論文の10件中1~10を表示しています
    • |<
    • <
    • 1
    • >
    • >|
論文
  • 原 一夫, 新保 仁, 松本 裕治
    22 巻 (2007) 3 号 p. 248-255
    公開日: 2007/02/01
    ジャーナル フリー
  • 参沢 匡将, 渡邊 恭子, 下川 哲矢
    22 巻 (2007) 3 号 p. 256-262
    公開日: 2007/02/15
    ジャーナル フリー
    Recently, we proposed an agent-based model called the word of mouth model to analyze the influence of an information transmission process to price formation in financial markets. Especially, the short-term predictability of asset return was focused on and an explanation in the view of information transmission was provided to the question why the predictability was much clearly observed in the small-sized stocks. This paper, to extend the previous study, demonstrates that the word of mouth model also has a consistency with other important financial stylized facts. This strengthens the possibility that the information transmission among investors plays a crucial role in price formation. Concretely, this paper addresses two famous statistical features of returns; the leptokurtic distribution of return and the autocorrelation of return volatility. The reasons why these statistical facts receive especial attentions of researchers among financial stylized facts are their statistical robustness and practical importance, such as the applications to the derivative pricing problems.
    抄録全体を表示
  • 武内 雅宇, 小路 悠介, 來村 徳信, 林 雄介, 池田 満, 溝口 理一郎
    22 巻 (2007) 3 号 p. 263-275
    公開日: 2007/02/27
    ジャーナル フリー
    It has been recognized that design rationale is necessary for understanding and designing artifacts. In order to capture it, we focus on functional knowledge as design rationale and have developed a functional knowledge modeling tool, named SOFAST, which has been deployed in a few companies. In this paper, we aim to capture the processes related to functional knowledge which are not dealt with by the current SOFAST. We discuss a systematic description and an automatic capture system of two typical processes: modification process of functional decomposition trees in the context of designer's daily jobs, and systematization processes in which functional knowledge grows. For the implementation of the system, we extended SOFAST using an ontology-based Knowledge Management framework. Through demonstration and trial use of the system, SOFAST's users positively evaluated the system's effectiveness and expectable usefulness for their daily work.
    抄録全体を表示
  • 岡田 将吾, 伊藤 芳子, 長谷川 修
    22 巻 (2007) 3 号 p. 276-290
    公開日: 2007/03/20
    ジャーナル フリー
    In this research, we proposed multi-robot system that has robustness to the change of environments and has adaptability to the change of tasks and number of robots. To implement them, we use parameter learning on subsumption architecture.The subsumption architecture control robots' fundamental behaviors and robots move continuously in principle.The optimal value of the parameter and the number of robot that does the task in field are searched by learning. As the result,the robot's action on the task is improved (updated).Proposed system has functions to recognize the task and to generate the recordation memory.And then robots adapt to the change of tasks and the number of robots.In this paper, the function of proposed system is tested in simulation experiments.
    抄録全体を表示
  • 池ヶ谷 有希, 野口 靖浩, 小暮 悟, 伊藤 敏彦, 小西 達裕, 近藤 真, 麻生 英樹, 高木 朗, 伊東 幸宏
    22 巻 (2007) 3 号 p. 291-310
    公開日: 2007/03/20
    ジャーナル フリー
    This paper describes how to perform syntactic parsing and semantic analysis in a dialog system. The paper especially deals with how to disambiguate potentially ambiguous sentences using the contextual information. Although syntactic parsing and semantic analysis are often studied independently of each other, correct parsing of a sentence often requires the semantic information on the input and/or the contextual information prior to the input. Accordingly, we merge syntactic parsing with semantic analysis, which enables syntactic parsing taking advantage of the semantic content of an input and its context. One of the biggest problems of semantic analysis is how to interpret dependency structures. We employ a framework for semantic representations that circumvents the problem. Within the framework, the meaning of any predicate is converted into a semantic representation which only permits a single type of predicate: an identifying predicate "aru". The semantic representations are expressed as sets of "attribute-value" pairs, and those semantic representations are stored in the context information. Our system disambiguates syntactic/semantic ambiguities of inputs referring to the attribute-value pairs in the context information. We have experimentally confirmed the effectiveness of our approach; specifically, the experiment confirmed high accuracy of parsing and correctness of generated semantic representations.
    抄録全体を表示
  • 安藤 晋, 佐久間 淳, 鈴木 英之進, 小林 重信
    22 巻 (2007) 3 号 p. 311-321
    公開日: 2007/03/20
    ジャーナル フリー
    Unsupervised learning techniques, e.g. clustering, is useful for obtaining a summary of a dataset. However, its application to large databases can be computationally expensive. Alternatively, useful information can also be retrieved from its subsets in a more efficient yet effective manner. This paper addresses the problem of finding a small subset of minority instances whose distribution significantly differs from that of the majority. Generally, such a subset can substantially overlap with the majority, which is problematic for conventional estimation of distribution. This paper proposes a new approach for estimating a minority distribution based on Information Theoretic Framework, an extension of the Rate Distortion Theory for unsupervised learning tasks. Specifically, the proposed method (a) estimates parameters which maximize the divergence between the minority and majority distributions, (b) penalizes the redundancy of data expression based on the mutual information between the observed and hidden variables, and (c) employs a hard assignment approximation to avoid computation of trivial conditional probabilities. The algorithm of the proposed method has no problem-dependent parameter and its time and space complexities are linear to the size of the minority subset. Experiments using artificial datasets show the proposed method yields significantly high precision and sensitivity in detecting minority subsets which substantially overlaps with the majority. The proposed method also substantially outperforms one-class classification and mixture estimation methods in real-world benchmark datasets for text and satellite imagery classification.
    抄録全体を表示
  • 若木 利子, 沢村 一, 福本 太郎, 向井 孝徳, 新田 克己
    22 巻 (2007) 3 号 p. 322-331
    公開日: 2007/04/03
    ジャーナル フリー
    Though many kinds of multi-agent systems based on argumentation have been proposed where only rule-based knowledge is taken into account, they have been unable to handle the ontological knowledge so far. In our daily life, however, there are a lot of human argumentation where both ontological and rule knowledges are used. For example, in e-commerce, a seller and a buyer usually use ontologies about products along with their respective strategic rules for buying and selling. Recent progress of the Semantic Web technology provides expressive ontology languages. In this paper, we demonstrate integration of the Semantic Web reasoning and argument-based reasoning. We have implemented the integrated system such that Logic of Multiple-valued Argumentation-based agent system (specialized to two values {f, t }) can be accessible to the Semantic Web reasoning established as the description logic reasoning system, given ontologies expressed by OWL DL or its notational variant the DL SHOIN(D). An interesting argumentation result using both ontologies and rules about the university curriculum is shown as an example executed by our system.
    抄録全体を表示
  • 宮崎 和光, 木村 元, 小林 重信
    22 巻 (2007) 3 号 p. 332-341
    公開日: 2007/04/03
    ジャーナル フリー
    Reinforcement Learning is a kind of machine learning. We know Profit Sharing, the Rational Policy Making algorithm (RPM), the Penalty Avoiding Rational Policy Making algorithm and PS-r* to guarantee the rationality in a typical class of the Partially Observable Markov Decision Processes. However they cannot treat continuous state spaces. In this paper, we present a solution to adapt them in continuous state spaces. We give RPM a mechanism to treat continuous state spaces in the environment that has the same type of a reward. We show the effectiveness of the proposed method in numerical examples.
    抄録全体を表示
  • 矢入 健久, 前野 俊昭
    22 巻 (2007) 3 号 p. 342-352
    公開日: 2007/04/03
    ジャーナル フリー
    Covisibility-based mapping is a paradigm for robotic map building research in which a mobile robot estimates multiple object positions only from ``covisibility'' information, i.e., ``which objects were recognized at a time''. In previous studies on this problem, a solution based on a combination of heuristics - ``closely located objects are likely to be seen simultaneously more often than distant objects'' and Multi-Dimensional Scaling (MDS) was proposed, and it was shown that qualitative spatial relationships among objects are learned with high accuracy by this method. However, theoretical validity of the heuristics has not been sufficiently discussed in these studies. Besides, the existing method has a defect that the quantitative accuracy of built maps is very low. In this paper, we first prove that the heuristics is generally valid in a certain condition, and then present several enhancements to the original method in order to improve the quantitative accuracy of the maps. In the experiments, it was found these enhacements are quite effective.
    抄録全体を表示
  • 矢入 健久
    22 巻 (2007) 3 号 p. 353-363
    公開日: 2007/04/03
    ジャーナル フリー
    In recent years, simultaneous localization and mapping (SLAM) based on stochastic state-transition / observation models and Bayesian estimation technique has been the mainstream of the mobile robot mapping research. In contrast to this trend, we present an alternative formulation of the map building problem from the viewpoint of non-linear dimensionality reduction or manifold learning. In this framework, the robot map building is interpreted as a problem of reconstructing the coordinates of objects so that proximities between them in the space of robot's observation history as faithfully as possible. Based on this insight, we generalize the covisibility-based mapping method which was established in previous studies into the map building based on dimensionality reduction of historical visibility data. We applied latest non-linear dimensionality reduction techniques to this framework, and compared them with classical techniques such as PCA and MDS in experimental studies.
    抄録全体を表示
    • |<
    • <
    • 1
    • >
    • >|
feedback
Top