[in Japanese]

View full abstract

[in Japanese]

View full abstract

This paper deals with the perceptual system for the perception-based robotics. The robotic system is composed of the perceptual system and the action system. The fuzzy controller is adopted for the action system. Genetic programming (GP) is adopted to generate analytic functions associated with a perceptual system of the mobile robot. These functions extract effective features from a set of sensing data. The robot can acquire its perceptual system which is suitable to the facing environment by generating functions to translate sensing information to perceptual information. The effectiveness of the proposal approach is demonstrated by computer experiments for collision avoidance problem.

View full abstract

Fuzzy Flexible Interpretive Structural Modeling (FISM/fuzzy) is a fuzzy version of Flexible Interpretive Structural Modeling (FISM). In this paper, we propose a consensus process to develop a fuzzy consensus model from multi user's fuzzy models. The consensus process consists of two sub-processes; compare process and compromise process. Compare process in to obtain the group difference using a comparable model of the structure from multi user's models. A fuzzy comparison matrix to compare fuzzy reachability matrics is proposed. Compromise process is to construct a fuzzy structure model from a comparable model with discussions of users. We discuss compromise strategies to construct a fuzzy matrix that most users accept.

View full abstract

In this paper, focusing on multiobjective two-level 0 1 programming problems where two decision makers with 0-1 decision variables have multiple objective functions, we present interactive fuzzy programming for them using partial information about preference. In the proposed method, after introducing fuzzy goals of objective functions for the decision makers, a solution, which minimizes the difference between the membership function values and the reference membership levels for the decision maker at the lower level under the condition that the membership is obtained by genetic algorithms. After evaluating the aggregated satisfactory degree for each of the decision makers through the partial information about preference, if the ratio of the aggregated satisfactory degree for the decision maker at the lower level to that for the decision maker at the upper level satisfies a certain condition and the decision maker at the upper level is content with the solution, a satisfactory solution is obtained. Otherwise, the interactive procedures with updating the minimal satisfactory levels and the reference membership levels is going to be continued until a satisfactory solution is found.

View full abstract

This paper proposes an acceleration method of GA search that finds a new elite by fitting a single-peak function on fitness landscape. The roughest approximation of a finite fitness landscape that has one global optimum would be a single-peak curved surface, and the vertex of the approximated single-peak function is expected to be near the global optimum of the original searching space. We propose two data selection methods for the fitting, use a quadratic function as the single-peak function, and evaluate the proposed idea using seven benchmark functions. The experimental results have shown that the proposed method accelerate GA convergence.

View full abstract

In this paper, we propose an 4-step inference method which is necessary to construct natural language communication system. The method is referred to how to estimate Fuzzy Quantifier Q''' when QA is F is γ ⇔ Q'''(m'A) is mF is m''γ is inferred (Q, Q''':Qquantifiers, A: Fuzzy subject, m, m', m'':Modifier, F:Fuzzy predicate, γ:Truth-qualifier). We showed the Q''' is resolved by 4-step procedure for two types of Q. These two types include a nonincreasing type (FEW,...) and a nondecreasingtype (MOST,...). First step we infer QA is F is γ ⇔ Q''(m'A) is F is γ, we transform A → m'A, and get Q'. Second step we infer ''Q'(m'A) is F is γ ⇔ Q''(m'A) is mF is γ, we transform F → mF, and get Q''. Third step we infer Q''(m'A) is mF is γ ⇔ Q(m'A) is mF is TT, we transform Q'' → Q, and get truth-qualifier TT. Forth step we infer Q(m'A) is mF is TT ⇔ Q'''(m'A) is mF is m''γ, we transform truth-qualifier TT → m''γ, and get Q'''. By above 4-step, we get the inference result as a fuzzy quantifier in case of transforming simultaneously a fuzzy subject and a fuzzy predicate and a truth-qualifier.

View full abstract

In this paper, a method of extracting local minor components, that takes memberships to cluster into account, is presented. A simple neural PCA approach is employed to extract the components. The components well represent the distinctive structures of the data set. The way of estimating missing values by using the local minor components is also introduced.

View full abstract

In object-oriented software development, class design is one of the most significant factors. In order to provide assessments or predictions of class designs, software metrics have been studied in software engineering discipline. When a software metric does not give a reasonable evaluation to a class, we could modify the class structure, and our modification could lead to simplify the class design quality. In order to support such modifications in terms of class structural complexity, an analysis-method with fuzzy graph has been proposed in the past. This is a method for detecting a class member -a method or an attribute- such that it may have a significant impact on the class structural complexity. However, the method effectiveness has never been discussed. The contribution of this paper is to present an empirical study on the effectiveness of that method. In the empirical study, many object-oriented programs (Java classes) are made pseudo modifications, and impacts of those modifications are evaluated by a software metric. The empirical results show that the class members corresponding to the strengthen points in terms of unilaterally connected graph or strongly connected graph could have significant impacts on the class structural complexity. Thus, the above method is useful for class structural analysis.

View full abstract