This paper proposes a new method that analyzes textual data by using key phrase patterns, where the patterns extracts key phrases and the key phrases give features composed of both words and parts of speech to each item of textual data. The patterns are created by using linguistic knowledge but without using knowledge of a task. The patterns are able to be applied to many tasks. The method applies lexical analysis to items of textual data and extracts key phrases using key phrase patterns. Also, the method uses a fuzzy inductive learning algorithm and acquires relationships between key phrases and text classes of the items given by a user. The acquired relationships give new knowledge to user and realize to infer a text class corresponding to a new item of textual data. This paper applies the method to two kinds of e-mail analysis tasks, where the e-mails are collected by our customer center. The results based on the tasks indicate that the method does more valid analysis than the method based on words.

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

This paper considers multiobjective linear programming problems involving fuzzy random variable coefficients in the objective functions. A novel decision making model, which incorporates the concept of α-level set and a fractile criterion optimization model in stochastic programming, is proposed. By extending M-α-Pareto optimal solutions in fuzzy programming, a solution concept in the proposed model is defined. In order to derive a satisficing solution for a decision maker among a set of the defined solutions, an interactive satisficing method is constructed. Finally, a simple example is provided.

View full abstract

This paper aims at discussing a clustering algorithm based on Learning Vector Quantization (LVQ) using a kernel function in support vector machines. Mapping object data into the high-dimensional feature space, this algorithm can find nonlinear boundaries between clusters which ordinary algorithms cannot find. The reason why kernel-based algorithms can find nonlinear clusters is that they may be linearly separated in the high-dimensional feature space. Nevertheless, actual configuration of data units in the high-dimensional feature space is unknown. Self-Organizing Map (SOM) associated with LVQ is hence applied with a kernel function. The resulting topological map for data in the high-dimensional feature space can visualize linearly separated clusters found by the proposed method. Numerical examples are given to show effectiveness of the proposed method when compared with fuzzy c-means and kernel-based fuzzy c-means.

View full abstract

In this paper, we propose a subjective visualization method for reading the environment of the exchange rate based on an optimistic-pessimistic axis by intuitive reasonings. The intuitive reasonings are totally different from logical reasonings used in AI and are defined by subjective information.Firstly, the intuitive reasonings are briefly described. Secondly, intuitive reasonings using Neural Networks are applied to the exchange rate to give us two subjective evaluations on dealing change width and dealing change direction. Thirdly, these subjective exchange rate evaluations and their certainty factors are projected on the optimistic-pessimistic axis and subjective environment maps of the exchange rate is created. Finally, some typical subjective environment maps according to the actual data of the domestic exchange are shown and the authors read the environment of the exchange rate based on the subjective maps.

View full abstract

In this paper, we propose a new reinforcement learning algorithm with a variable bias between two animalistic instincts-the activeness and the cautiousness. Although reinforcement learning methods are more flexible than supervised learning methods, it is difficult to determine the appropriate amount of reinforcement signals. For example, too much reward for the second best solutions can prevent the agent from exploring for the best solution, while insufficient amount of reward can prevent it from learning any kind of solutions. To overcome such problems, the proposed model uses two learning modules with a variable bias. One of the modules represents the activeness. Its sole goal is to maximize the amount of reward given from the environment. The other module represents the cautiousness. Its goal is to minimize the amount of penalty. By changing the bias between these two modules, the proposed model can perform efficient learning in wider variety of environments. With computer simulations, we confirmed that the proposed model can learn effectively in environments where conventional models show poor performance. In other environments, the proposed model showed performances comparable to conventional models.

View full abstract

Recently, human-friendly robots are expected to support human daily life at home, office, and scene of medical treatment and welfare with rapid declining population of children and aging problem. This research purposes to develop the guide robot that can be used in hospitals and welfare facilities etc. Concretely, in order to achieve the same performance as human guidance on a robot, first, the advanced guidance knowledge is formulated using production rules based on linguistic variables in this paper. Secondarily, a new trajectory planning method for guide robots is proposed based on quantified knowledge and distance-type fuzzy reasoning. The distance-type fuzzy reasoning method can be applied, when the common set of an antecedent part and a fact is an empty set, because this resonning uses the distance value between two fuzzy sets. Finally, the effectiveness of the proposed trajectory planning method is shown by the experiment.

View full abstract

This paper proposes a distance-type diagrammatic reasoning method based on the distance information between figures. In the knowledge representation within the brain, visual figure information is figurative, and since there is much amount of information, not only symbolic information but also figure information is very important. Therefore, the knowledge representation with figures is often used for reasoning. In order to realize more humanlike reasoning, in this paper, the conceptual expression in the feature space is discussed first, and then the new reasoning method based on the distance information between figures is stated using the correspondence relation between the figure and the vector in the feature space. Based on this reasoning method, the figure reasoning system which can operate on Windows is developed. Conducting a simulation experiment using this reasoning system shows the feature of the distance type figure reasoning method.

View full abstract