A multiattribute utility function can be represented by a function of single-attribute utility functions if the decision maker's preference satisfies additive independence or mutually utility independence. Additive independence is a preference condition stronger than mutually utility independence. The multiattribute utility function is represented in the additive form if the former is satisfied and it is in the multiplicative form if the latter is satisfied. In this paper, we present methods for sensitivity analysis of multiattribute utility functions in the multiplicative form and also propose methods incorporating fuzziness of the decision maker's judgments in the procedures for assessing scaling constants (attribute weights).

View full abstract

It is very difficult to realize medical diagnosis systems on a computer, because in order to judge many sick support groups synthetically, generally one needs to use the data that consists of dozens or more items (the amount of the features). In this paper, a computer-aided medical diagnostic system is constructed using the Three-valued Cellular Neural Network (T-CNN), which has achieved excellent results in pattern classification, and the classification capability is examined by using liver disease data as an example. Consequently, a comparatively good diagnostic result is obtained from the T-CNN under the conditions examined by this study. Because the T-CNN can realize a system which performs the judging comparatively near humans, and tends to utilize the features of the whole sickness pattern, it was concluded that it could perform judgment nearest to the overall judgment by a specialist.

View full abstract

Fuzzy morphology has been proposed as a new method of image processing, especially the analysis of characters from an ambiguous image. Fuzzy morphological operators work with two images: an original image to be processed and a structuring element. Generally, we must decide on the shape and value of the structuring element before applying fuzzy morphology. A problem arises when the applied image changes largely by the selection of the structuring element, so it is difficult to apply fuzzy morphological operators to an automatic inspection system for defect testing. In this paper, we propose a new structuring element for fuzzy morphology, called a fuzzy adaptive structuring element. The fuzzy adaptive structuring element (FASE) determines shapes and values of structuring elements dynamically from an input image by searching for the local regions in the image. Consequently, the fuzzy adaptive structuring element is sensitive to ambiguous images, useful for automatic inspection using fuzzy morphology, and effective for extraction. Furthermore, the fuzzy adaptive structuring element shows good performance resulting from a package defect testing compared with the other filters. We have got experimental results (more than 95.65%) by applying a fuzzy adaptive structuring element to seal defect testing.

View full abstract

This paper discusses time expressions as clues to understand a user's intention. We employed three questionnaires to find relationships between hearers' guesses on semantically ambiguous phrases and time expressions in them. The results showed that the questionnaires' respondents tended to change their guess corresponding to the granularity of the time expression in the phrase, and that this tendency is common, regardless of age or sex. The results also showed that almost all of the guesses could be classified into three categories: speaker's attribute, current state, and purpose or reason for action. Therefore, we concluded that it is effective to focus time expressions to clarify ambiguous utterances and understand intention.

View full abstract

This paper proposes a new fuzzy reasoning system with an entropy term as a regularization term in the objective function. Many studies have been done with fuzzy inference systems so far. Further, many improved models have been also proposed such as models using GA and self-organizing neural networks. The proposed fuzzy reasoning system has the objective function composed of the squared error and entropy terms: The former is the conventional objective one for the fuzzy reasoning system and the latter is a term which controls a shift of membership function. By adding the latter, learning of the system proceeds as degree of fitness of each rule approaches to 0.5. In order to demonstrate the validity of the proposed method, some numerical simulations are performed.

View full abstract