Kohonen's Self-Organizing Map (SOM) is a kind of neural networks that learns the feature of multi input data without supervision. Therefore, it seems difficult to apply the SOM, in the case that the feature of input data changes from learning example data. In this paper, the authors proposed that "Automatic Incremental Learning System of Self-Organizing Map" is effective for its solution. This system consists of two techniques. The one works automatic learning of the input data feature, when the input data has partly been shifted from features of learning example, and the other is "semi-automatic labeling" on new SOM feature map that adjust territory labels by using information from the former map.

View full abstract

This paper proposes a dynamic deletion method of units for SOM (Self-Organizing Map) by using the updated information of each unit on the competing layer. Concretely, during the process of learning, a unit with a low contribution to learning, is removed dynamically by each unit's ability to judge for itself (extinction of unit). This process may imitate a cell death in a living body. This dynamic deletion of units enables SOM to map the input data on the competing layer accurately with a small number of units. Furthermore, by using the proposed method, the input space with discontinuous distribution of data can be learned correctly.

View full abstract

In substations of the electric power systems, it is necessary to predict load power change accurately to improve their operating efficiency and reliability. This paper investigates an application of the Self-Organizing Map (SOM), an effective technique for classification of multi-dimensional data, to the prediction of load power change at the local substations. The SOMs are constructed by learning data of apparent, effective and reactive powers. Using the SOMs, load power patterns during a day are predicted well separately. To improve the prediction accuracy, the date of maximum and minimum atmospheric temperature is added to the above data of powers. In a short-term prediction, the powers during each four hours in a day are predicted by the SOM learned by the data during each 12 hours This paper also proves the effectiveness of our method with numerical results.

View full abstract

The bus voltage of local substations in the power system has been controlled so as to be maintained within the regulated range by reactive power control of the phase modifying equipments (PME) [1]〜[3]. In such a control, the switching operation of the PME sometimes cannot be implemented suitably even in a vicinity of the upper and lower limits of the regulated range of the bus voltage. And this causes unnecessary reactive power. As a result, the reactive power increases, especially in a long distance transmission, and hence the transmission loss increases. Some of the authors have investigated the optimal control of the PME by using the fuzzy reasoning [4] for maintaining the bus voltage within the desired range and at the same time for reducing the transmission loss through suppression of the generation of the unnecessary reactive power [5]〜[8]. In the control, since the fuzzy rules were constructed by using bus voltage and reactive power obtained in the past, when the load power changes due to the change in weather condition, the resultant rules become unsuitable, and hence sometimes unexpected control of the PME took place [7]. In the present paper, by using the self-organizing map (SOM) [11]〜[13] which is known as an effective technique to the classification of multi-dimensional data, changes of the bus voltage and the load power arepredicted. And these predicted results are applied to tuning of the fuzzy rules with an aid of the steepest descent method. Finally, the control of the PME corresponding to the change of the load power can be realized. The effectiveness of our method is proved with numerical results by using real data in a power system.

View full abstract

Ultrasonic Motors (USM) possesses variable dead-zone in the control input associated with applied load torque. The dead-zone is a problem as an accurate positioning actuator for industrial applications and it is important to eliminate the dead-zone in order to improve the control performance. For the purpose of solving the problem, this paper presents a position control scheme with fuzzy dead-zone compensation for ultrasonic motors. In the proposed control scheme, fuzzy inference is adopted to determine a dead-zone compensation input and construct a plant which compensated the influence of dead-zone. The effectiveness and validity of the proposed controller is demonstrated by experiments.

View full abstract

A floor plan in architecture includes elements that determine circulations of a person who experiences the space of the architecture and that is considered to play an important role in the connection between the person and the architecture. Especially since modern times, various shapes that deviate and escape from a floor plan by means of right-angled constitution have arisen and have been used as elements of architecture. Therefore, it is necessary to consider "a floor plan". In this paper, we propose an index "Grade of surprise" that is able to classify various floor plans in architecture from the person's view. We analyze floor plans of architecture by famous architects as some samples by means of a proposed "modified box-count method". We compare the results and make consideration.

View full abstract

In this paper, a fuzzy number on the real line is defined as the one whose membership function is everywhere continuous and has a strict monotonicity on its bounded support. Firstly, it is proved that the family of all fuzzy numbers is closed under the operation of multiplication derived from Zadeh's extension principle. Two kinds of subfamilies of fuzzy numbers closed under the multiplication are presented. For the purpose of describing these two subfamilies, the notion of polynomial type and piecewise-polynomial type fuzzy numbers are introduced. Fundamental formulae for the multiplication among fuzzy numbers of polynomial type or piecewise-polynomial type are given.

View full abstract

The pregnancy will take place in a short term a rapid change, and he one of the period which falls easily into a psychologically critical situation in woman's lifestyle. The social support from others is a method to decrease psychological anxiety. In this paper, the analysis of relationship between pregnant woman's anxiety and the social support was analyzed by using statistical analysis and fuzzy structural modeling "FISM/fuzzy". It was able to make comparative study of the structure of the social support in the pregnant woman with different level of anxiety by using FISM/fuzzy, and the analysis to relax anxiety concerning the social support became possible.

View full abstract

In this paper, we proposed a multistage fuzzy inference system consisted of several two-input one-output fuzzy inference engine, which infers and predicts values of a set of time-series data. The inference engine use a two dimensional fuzzy rule array to represent the time-series data production mechanism. To construct the rule arrays, we extract the knowledge of production from a real time-series data by genetic algorithm. We applied the method to extract the anesthetist's knowledge for blood pressure control in surgical operation. It was found that it is useful for data-mining to acquire the knowledge about the time-series data, automatically. The acquired knowledge as rule arrays are not sufficient because the number of time-series data available is quite restricted and these subjects needs to be investigated for future problems.

View full abstract

In this paper, the application to motion evaluation of interval regression analysis is described. The analysis method about motion environment was described first, next motion evaluation by interval regression analysis was performed. The result was examined about applicability over motion evaluation of interval regression analysis and variation in evaluation to a stimulus as compared with the result of method of equal appearing intervals. Furthermore, the result of interval regression analysis was calculated for the relation between permissible value of motion and subjectivity evaluation as compared with permissible value of the motion environment by ISO.

View full abstract

Human face-to-face interaction consists of both verbal and nonverbal communication such as facial expressions. Therefore many researchers have used facial expressions for human and computer interaction in order to realize a pleasant interaction with a computer system. However, there have been few researches about what kinds of faces should be suitable for human interfaces in terms of facial affect recognition. We conduct an experiment in which 30 subjects are asked to identify an upright face with a rotated one. The results indicate that the faces which express some affect clearly could be recognized more rapidly than the other ones. Facial expressions appropriate for human computer interfaces are discussed on the basis of the results.

View full abstract