This paper proposes a method for diagnosis of glaucoma by using Learning Vector Quantization(LVQ) technique from a fundus image. The diagnosis accuracy was determined by using the proposed method from the data set which consists of four components. They are two size components and two threshold ones in the excavation of the optic disk and the optic disk. In the experiments, we have achieved a maximum accuracy rate of 78.9%, which is comparable to the results by the observer appraisals of the fundus images.

View full abstract

This paper reports on practical examples for algorithm education using colors to visualize sort algorithms. Colors gradation is effective for understanding of Sort algorithms' features. Visual methods for algorithm education are proposed and examples for visualization of sort algorithms using colors are presented.

View full abstract

In order to solve the problems of ill education due to student comprehensive ability's wide gap in school, the education improvement for group of students with similar ability is expected. Thus, the paper proposes the application of the similar test result student retrieval system to the above grouping in the information literacy education.

View full abstract

Observation pictures of botanic plant growth in the natural environment change have been stored together with the camera shot day's weather data for education. Botanic plant growth state in the picture has been categorized in either of standard plant growth states with membership-function grade value so that the database of the botanic plant growth state and weather data can be constructed for basket analysis of plant growth rate balance change in the same camera-shot day's pictures. Weather data has been expressed by the sound, which consists of the tone sound part and the up-down sound part. The former part shows the usual weather data by tone. The latter part shows the difference of camera shot day's weather data and the usual weather data. The sound is used to check whether or not any abnormal state can be discovered in each month botanic plant observation pictures' set and plant growth rate association check tables.

View full abstract

We have been researching about the transformation from Picture to Music. There are special characteristics such as Fractal Dimension and Zipf's law. Some of Turner's oil paintings gave moderate impression, but the changes of color were great. We analyzed his picture and composed music that might represent the special feature of his picture.

View full abstract

This paper proposes the graph expression and analysis approach for sound design, so that the information engineers can compose the harmonious sound easily. Optimality grade of multiple part sound sequence in the fuzzy allowable sound sequence set will be derived by the graphic analysis.

View full abstract

Recently, the use of three dimension map is increasing due to spreads of the car navigation system and the streetscape simulation, etc. A lot of processes of the hand work, however, are included in making such 3D map under the present situation, and it is a current state to require a large labor. Therefore, these processes are automated, and the system for making the 3D map efficiently is expected. In this study, we propose a method for parsing the streetscape image as a preprocessing to make the 3D map by stereo method. In this report, a hierarchical method that uses an image pyramid and a fuzzy inference is proposed and is applied to the streetscape image. As a result of experiment, we confirmed the effectiveness of the proposed method as compared with an ordinal method.

View full abstract

An objects detection algorithm for color dynamic images from two cameras is proposed under low illumination for a real surveillance system. It provides automatic calculation of a Fuzzy Corresponding Map and color similarity for lower luminance conditions, which detects little chromatic regions in input images under lower illumination taking perspective into consideration. Experimental detection results for two dynamic images from real surveillance cameras in a downtown area in Japan under low illumination show that the proposed algorithm has 15% improved accuracy compared with the independent detection algorithm in the same false alarm rate. The proposed will be used as a basis unit for surveillance camera system.

View full abstract

In this paper, we propose a method for generation of stained-glass-style images based on a Voronoi diagram. In the proposed method, Voronoi generators are re-arranged according to the differences between an original image and the image represented by using the Voronoi diagram. The effectiveness of the proposed method is verified by applying it to some digital images.

View full abstract

This paper proposes the generation system of Japanese riddles using variety of verbal meanings between homonyms. The system deals with interesting riddles that require verbal semantic information of causal relation. The proposed system consists of the generation section and two databases. The first database has verbs, their using frequencies, and subjects or objects for verbs. The second database has effects caused by actions expressed by verbs and the strength of causal relation between actions and effects. The generation section generates interesting riddles using their databases and based on the verbs using frequencies and the strength of causal relation. This paper also shows subjects experiments in order to confirm the validity of riddles generated by the proposed system.

View full abstract

There exist many problems in designing a human-to-human communication medium though people recognize the significance of medium as a communication tool. The ''affordance'' theory provided an ecological alternative to cognitive approaches and the concept is popular in the field of user interface design. We extend the concept to cover human-to-human communication and propose the novel concept of ''human affordance,'' which is afforded from humans, not artifacts. We describe the significance of affordance augmentation which creates or enhances mutual understanding in communication by media processing technologies. Also, we introduce two case studies of human affordance; value sharing in text-based communication focused on evaluation differences of the strength of agreement / disagreement phrase, and Electronic Yosegaki (E-Yosegaki) as an N-to-1 communication tool, which enables message communication by hand writing from multiple people to an individual on a tablet computer.

View full abstract

Recently, biometric personal identification is evolving in the field of security and personalized services. Personal identification based on walking pattern using load cell, camera and pressure sensor is human-friendly system. The pressure sensor in mat can obtain information, such as foot pressure and foot size. In this study we propose a personal identification method by pressure change in sole. In our method, we divide pressure change in sole for the features, such as peak pressure, timing of heel strike and timing of toe push off. We consider two types of personal identification method: Euclidean distance based method and the other is neural network. As the result, the method of Euclidean distance could identify one of 10 subjects at 70% recognition ratio at 4 divisions The method of neural network could identify one of 10 subjects at 85% recognition ratio at 5 divisions.

View full abstract

In order to reduce background noise in noisy speech, we have investigated a noise reduction method based on a noise reconstruction system (NRS). The NRS uses a linear prediction error filter (LPEF) and a noise reconstruction filter (NRF). An input signal of a LPEF becomes a white signal. Assuming that background noise is generated by exciting a linear system with a white signal, the background noise can be reconstructed from white noise by estimating the linear system. It is a NRF based as system identification that estimates the linear system. However, in case a fixed step size for updating tap coefficients of a NRF is used, it is difficult to reduce the background noise while maintaining the high quality of enhanced speech. Therefore, a variable step size for normalized least mean square (NLMS) is proposed in this paper. In a speech section, a small step size is used so as not to estimate speech, while a large step size is used to track the background noise in a non-speech section.

View full abstract

This paper describes an ultrasonic registration system for locating screw hole positions of intramedullary nail. Currently, X-ray has been used for visualizing the nail into the femur. However, it has serious problem of X-ray exposure. We propose a locating method of the screw hole positions by doing a image registration. We perform an experiment by one-direction manual scanning using an ultrasonic array probe. In our method, first, we extract screw hole regions by calculating two fuzzy degrees: average of the intensity of eight neighborhood and variance of the intensity along long axis of the nail using fuzzy inference. Next, we do a registration between the obtained image with the exact image, where the exact image is the truth image of the nail. The position of screw holes in the obtained image is calculated by moving from coordinate of screw holes in the exact image using the translation matrix. As the result, we could locate the screw holes on center axis of the nail and calculate the center distance of two screw holes within an error of 1.0 mm.

View full abstract

In the previous research developmental change in various functions of infants and children were examined. But there were few available data about developmental change in those functions of the adult period. Therefore, this investigation was designed to examine those problems. College student subjects were asked to evaluate 20 items to see to what degree 1 year old, 3 year old, 5 year old, 10 year old, 20 year old, 40 year old, 60 year old and 80 year old people would be able to perform independently and adequately. The mean value of the 20 items increased when the age rises from 1 years old to 20 years old. The mean value of all items became the highest at 20 years old or 40 years old. For 60 years old and 80 years old, the mean value of all items decreased more than the mean value of 40 years old. Correlations with representative values calculated by different method (mean, max, min, etc.) of 5 item-groups and the independence level in each age are compared and discussed.

View full abstract

There are many numbers of studies that discuss algebraic structures on fuzzy truth values. But, their main interest is in convex fuzzy truth values. This paper concerns with non-convex fuzzy truth values. A multi-interval truth value is defied as a collection of interval truth values. Therefore, it is not convex in general. This paper focuses on the simplest multi-interval truth values, ternary multi-interval truth values. A ternary multi-interval truth value is a non-empty subset of {0, 1, 2}. Thus, multi-interval truth value {0, 2} is not convex. Min, Max, and involution are defined over multi-interval truth values by Zadeh's extension principle. Then, it is known that a set of multi-interval truth values with the three operations is a de Morgan bi-lattices. This means that non-convex fuzzy truth values do not form a lattice any more. This paper describes functions over ternary multi-interval truth values which are expressed by logic formulas. Then, the paper clarifies that the functions are determined by the four multi-interval truth values {0}, {1}, {2}, and {0, 2}.

View full abstract

We propose a simple calculation method of attribute reduction in variable precision rough set (for short, VPRS) models proposed by Ziarko. VPRS is an extension of Pawlak's rough set theory, and it provides a theoretical basis to treat inconsistent or probabilistic information in the framework of rough sets. Inuiguchi has proposed various types of attribute reduction in VPRS, and Wang and Chiou have proposed a method to calcurate reducts preserving beta-lower approximations (called L-beta-reducts) proposed by Inuiguchi. In this paper, we prove that calculation of L-beta-reducts results in calculation of relative reducts in Pawlak's rough set theory. Moreover, by generating a beta-decision table from the given decision table, we propose a method to calculate all L-beta-reducts in the given decision table by using a discernibility matrix.

View full abstract

A fuzzy (linguistic) truth value is a truth value specified by a fuzzy set over a closed interval [0; 1]. Logical operations, defined according to the extension principle, do not satisfy all identities to be lattice because there are subnormal truth values and nonconvex truth values that violate the absorption laws. In 2000, Brzozowski proposed de Morgan bisemilattice, which is generalized algebra of de Morgan lattice in order for applications in multi-valued simulations of digital circuits. This paper studies a notion of fuzzy-interval equivalnt relation defined for fuzzy functions taking linguistic truth value.

View full abstract

In this report, the ordinal sum theorem of semigroups is applied to construct logical operations for several fuzzy logics. The generalized form of ordinal sum for fuzzy logics on [0,1] is defined in order to express uniformly several families of logical operations, and then the conditions for the conjunction/disjunction-likeness or for the left/right-continuity are presented.

View full abstract

In this paper, we examine a logical representation of images by means of multi-rough sets. We introduce a level of granularization into a color space and then define approximations of images. Further we extend the idea to conditionals in images, which is expected to give a basis of image indexing and retrieval by images themselves.

View full abstract

We often use fuzzy decision in many decision-making tasks, however, these tasks are too complex to be understood quantitatively. So it is difficult to analyze decision-making by using fuzzy decision. In this paper, we would propose a new method of fuzzy decision applying AHP.

View full abstract

By Inaida et al. shows that fuzzy reasoning does not sattisy monotoness. In this paper, we will show why their example does not satisfy monotoness.

View full abstract

We have used the fuzzy reasoning method "Min-Max Gravity Method", "Product-Sum-Gravity Method" and "Simplified Fuzzy Reasoning" for the fuzzy control and the soft science. The monotonicity is the important point for applying fuzzy reasoning to these field. However,the numerical reasoning functions of these methods do not always hold the monotonicity condition. In this paper, by defining some fuzzy mathematical concepts, we would examine some conditions for the monotonicity of "Simplified Distance-type Reasoning Method".

View full abstract

The fuzzy rule calculation method based on Choquet integral is proposed. This method has lots of good prosperities such as monotonicity when the fuzzy rule is monotonicity. In this paper, we compare it with simplified fuzzy reasoning, min-max gravity method and product-sum gravity method.

View full abstract

Though various contents are provided through the internet recently, it is not easy to collect favorite contents among huge amounts of contents in terms of user's preference. In this paper, we focus on the collaborative filtering algorithm in the recommender system. We propose a modeling approach for preference similarity model in collaborative filtering. In our approach, the model is constructed through optimization of MAE(Mean Absolute Error). The model decides the weights of preference similarity from the value of correlation coefficient and the number of items and variance values of evaluation ratings of each person. Through numerical experiments compared with conventional correlation coefficient based approach using Movie Lens data, we discuss validness of the model using the evaluation variance.

View full abstract

One of useful approach to intuitive text mining is construction of text maps that characterize mutual relation among text documents composed of many keywords. In the text mining tasks, text documents are first preprocessed into numerical weights such as tf-idf weights by considering term frequency and inverse document frequency. Principal component analysis (PCA) is used for constructing low dimensional plots of multivariate data. It is, however, often difficult to extract meaningful features from a low dimensional text map when "interesting structure" is concealed by many nonsignificant keywords. This paper proposes to use a liner fuzzy clustering-based variable selection mechanism for selecting keywords that are useful for characterizing text documents. In the variable selection model, the absolute typicality of keywords is estimated based on a graded possibilistic approach. An experimental result with a famous Japanese novel "Kokoro" by Soseki Natsume demonstrates the characteristics of the new approach.

View full abstract

We propose a method to estimate the semantic orientation(positive or negative) of products reviews, in which a pair of noun and adjective are selected as the characteristic words to determine the semantic orientation of the reviews and the sentiment orientation scores of the review are calculated according to the phrase' s orientation. The words to change the effect of characteristic words for the semantic orientation, such as "not", "very" "quite", are considered in the propose algorithm which determine the semantic orientation. In order to evaluate the proposed classification method, two kinds of product reviews, a liquid crystal display and a MP3 player, which contain 1400 reviews, are classified to two categories (positive or negative).

View full abstract

As known as "Year 2007 Problem" in Japan, decrease of opportunities for expertise/skill transfer in organization is becoming a considerable social concern, and some urgent countermeasures to that problem are seriously requested, especially in manufacturing industries. This study investigates into an effective usage of knowledge bases storing a number of account records, in order to let them distributed and shared among community members. We mainly focus on extracting some latent but commonly shared analogous structures with respect to human errors out of incidents/accidents stored in the "Failure Database". A text-mining approach is applied to the documents stored in the database and analogous relationships among accidental cases are identified in spite of their superficial irrelevance. Those "latent but common" structures are visualized as a network, which provides the users with new cues for in-depth analysis of priorly experienced failures.

View full abstract

In this paper, we proposed the system that extracts only the track back to the blog related to the content of the original article from among the track back, and developed the system as extension of Firefox. The flow of this system is as follow. The proposed system traces the track backs, analyzes the original article and the track back articles, calculates the evaluation value of each article, and judges the tarack back relations by evaluation values. To evaluate the proposed system, we applied it to various blogs, and it was confirmed to be able to extract the track backs with the relation as a result of the experiment. Especially, the track back spam was able to be filtered. Because it is difficult to judge the relation between the article and the article.we extracted only storongly-relation in the our experiment, so the proposed system sometimes filtered the article wanted to read by the user. We will improve the extraction algorithm in the future.

View full abstract

It is very important to improve the nursing care quality in a medical field. Currently, nursing-care data are collected via WWW from many hospitals in Japan. The collected data are stored into a database. Some nursing-care experts evaluate the collected data to improve nursing care quality. However, it is difficult for experts to evaluate the data because of huge number of nursing-care data in the database. In order to reduce workloads to evaluate nursing-care data, we propose a neural network based classification system. We use standard three-layer feedforward neural networks with backpropagation type learning. First, we extract attribute values from texts written by nurses for generating numerical training data. And then, we train a neural network using the training data. From computer simulations, we show the effectiveness of our proposed system.

View full abstract

A fuzzy classifier based on fuzzy c-means (FCM) clustering has shown a decisive generalization ability in classification. The FCM classifier uses covariance structures to represent flexible shapes of clusters. Despite its effectiveness, the intense computation of covariance matrices is an impediment for classifying a set of high dimensional data. This paper proposes a way of directly handling high dimensional data in FCM clustering and classification. The proposed classifier outperforms well established relational classifier known as k-nearest neighbor (k-NN) on the benchmark set of COREL image collection, which was used by James Wang for tests of his Simplicity System.

View full abstract

In this paper, we propose a new boundary extraction method, which achieves a precise boundary extraction of a plaque in an intravascular ultrasound (IVUS) image. The proposed method uses not only the seed points information but also the information obtained from the IVUS image. In this method, the boundary to be extracted is approximated by a polynomial series. The coefficients of the series are determined by the least square method with weight of a ratio of the between-class to the within-class variance. The validity and the effectiveness of the proposed method have been confirmed by applying it to a practical boundary extraction problem of a plaque in the IVUS image.

View full abstract

Two clustering algorithms that handle data with tolerance are proposed. One is based on hard c-means while the other uses the learning vector quantization. The concept of the tolerance is included in both algorithms. First, the concept of tolerance which implies errors, ranges and the loss of attribute of data is described. Optimization problems that take the tolerance into account are formulated. Since Kuhn-Tucker conditions give a unique and explicit optimal solution, an alternate minimization algorithm and a learning algorithm are constructed. Moreover, effectiveness of the proposed algorithms is verified through numerical examples.

View full abstract

In this paper, we propose a novel k-nearest neighbor classification bringing in the information of observation space. The proposed method uses not only the feature vectors but also the information of where the target data is taken. The effectiveness of the proposed method is verified by applying it to the tissue classification problem of the intravascular ultrasound (IVUS) data.

View full abstract

In this paper, we discuss c-means clustering algorithms on the multinomial manifold. Data forms a Riemannian manifold with the Fisher information metric via the probabilistic mapping from datum to a probability distribution. For discrete data, the statistical manifold of the multinomial distribution is appropriate. In general, The euclidean distance is not appropriate on the manifold because the parameter space of the distribution is not flat. We apply the Kullback-Leibler (KL) divergence or the Hellinger distance as approximations of the geodesic distance to hard c-means and fuzzy c-means.

View full abstract

Recently, many researches on Brain Computer Interface (BCI), which records brain activities and applies them for controlling of a wheelchair, typing and so on, have been reported. The authors have proposed a thought recognition method using Electroencephalogram(EEG) which can be measured noninvasively, therefore places low burdens on subjects. The accuracy of the proposed method, however, has not reached adequate level for practical use. To solve this problem, the authors have proposed a method in which EEG is fed back to the subjects. This method helps subjects aware of the thoughts they are thinking and generate recognizable EEG. The authors think that alpha rhythm, which is one of the EEG features derived from the parieto-occipital cortex and is considered to be useless for BCIs because it is influenced by visual stimuli, can be useful for a BCI based on the proposed method which doesn't employ visual feedback but auditory one. An experiment, in this paper, showed that some subjects had got skills to control their alpha rhythms through auditory feedback training with their eyes closed.

View full abstract

To investigate the brain activity during human recalling system of the memory, the authors recorded electroencephalograms (EEGs) from one subject in recalling English sentences which had stored by the loci mnemonic system. The subject had stored almost three thousands English sentences for seven years by the loci mnemonic method. We displayed one number corresponding to one sentence on CRT.

View full abstract

We have already proposed "biomodeling system", in which the "top-down bio-processing" for sending actuator signals to robot from living neuronal network cultured on a 2-dimensional electrode arrays, and the "bottom-up robot-processing" for electrical stimulation to living neuronal network from robot are connected between neuronal network and robot. In this paper, we discuss two kinds of learning mechanism, which are plasticity learning of living neural network and adaptability learning of fuzzy logic using the tracking estimation of Khepera II in a straight lane. Our goal is reconstruction of the neuronal network, which can process "thinking" in the dissociated culture system.

View full abstract

Recently, physical assisting equipments using biological information have been developed for crippled elderly or physically challenged people. This study focuses on electroencepharogram (EEG) as a biological information and aims to control various machines using EEG. In electroencephalogrambased control, it is often hard to generate effective discrimination model for thinkings because of changing thinking/EEG as time advances. The purpose of this study is to acquire effective and stable discrimination model by training, then to analyze measurement sites, frequency bands and thinking ways effective for discrimination. This paper investigates a visual real-time feedback for the training. This paper discusses the discrimination rate and the transition of affected measurement sites through the experiment by 4 subjects.

View full abstract

To investigate the brain activity during human recognition of characters and symbols representing directional meaning, the authors recorded electroencephalograms (EEGs) from subjects in viewing four types of Kanji (Chinese characters being used currently in the Japanese language) and arrows presented on the CRT. Each of four characters or symbols means direction of upward, downward, leftward and rightward, respectively. Further, subjects were asked to make read the stimulus silently. Regardless of the directions, the reaction time was almost equal, when character or arrow was presented. EEGs were averaged in each stimulus type and direction, and event-related potentials (ERPs) were obtained. The equivalent current dipole source localization (ECDL) method was applied to these ERPs. After 300 ms from presentation of stimuli, ECDs were localized to areas related to the language, such as the left supramarginal gyrus, the Wernicke's area in the left middle temporal gyrus, the left angular gyrus and the left lingual gyrus. Then ECDs were localized to the Broca's area. With arrow symbols, ECDs were localized to areas related to the shape recognition, such as the right inferior or the right middle temporal gyrus. In both cases, ECDs were localized to areas related to working memory for the spatial perception.

View full abstract

In preparing for the experiment, the subject learned the short english sentences by heart. The subject consolidated the memory by mnemonics and mental rehearsals in advance. In the event related fMRI experiment, when a visual stimulus of an image was projected on a liquid crystal screen, the recalling task by the subject followed. This experiment block was repeated and the MRI signals were measured. The fuzzy c-means (FCM) classifier is applied to the fMRI signals for classification of brain activities. The MRI signals constitutes massive high dimensional data. The classifier based on the matrices in terms of inner products can handle high dimensional data. The results show a possible application to brain communications.

View full abstract

We develop a design support system for room lighting environment using interactive evolutionary computation. When many LED's with different lighting levels are distributed on wide ceiling and walls, the number of optimization solutions for desired lighting environment drastically increases. We introduce IEC into a lighting design support system to overcome this problem. Usually, it take time to render many CG room objects, which hinders lighting design support even IEC is used. We overcome this problem by introducing a fast rendering method that approximates real rendering. First, we make a 2-D room lighting image from a fixed view angle when only one LED is turned on. We made many similar 2-D images by changing the turned LED. The fast rendering method is to generate any lighting image by weighted sum of the 2-D images. By combining the IEC method and the fast rendering method, the lighting design support system can be used by not only professional but also any users practically.

View full abstract

Combining multiple classifiers into a single one is a promising approach to the design of reliable classifiers in pattern classification problems. A number of approaches have been proposed for generating ensemble classifiers. Some studies indicate that the diversity of component classifiers is important to generate ensemble classifiers with high generalization ability. In this paper, we examine the effectiveness of genetic rule selection with a multi-classifier coding scheme for ensemble classifier design. Genetic rule selection is a two-stage method. The first stage is rule extraction from numerical data using a data mining technique. Extracted rules are used as candidate rules. The second stage is evolutionary multiobjective rule selection from the candidate rules. We use a multi-classifier coding scheme where an ensemble classifier is represented by an integer string. Three criteria are used as objective functions in evolutionary multiobjective rule selection to optimize ensemble classifiers in terms of accuracy and diversity. We examine the performance of designed ensemble classifiers through computational experiments on six benchmark datasets in the UCI machine learning repository.

View full abstract

The search ability of different evolutionary multiobjective optimization algorithms is often compared using evaluation measures of obtained non-dominated solution sets. The hypervolume measure is one of the most frequently used evaluation measures in the literature. Recently several algorithms have been proposed to directly maximize the hypervolume measure. Those algorithms are often called indicator-based evolutionary algorithms (IBEAs). In our former study, we proposed an iterative IBEA where a single solution is obtained from its single run. In each run, our iterative IBEA searches for a solution with the maximum contribution to the hypervolume measure. In this paper, we slightly modify it to improve its search ability. In the modified version, first our iterative IBEA searches for an optimal solution of each objective. We do not use the hypervolume measure in this stage. After an optimal or near optimal solution of each objective is obtained, our iterative IBEA searches for a solution with the maximum contribution to the hypervolume in its each run. Through computational experiments on multiobjective 0/1 knapsack problems, we demonstrate that the simple modification significantly improves the search ability of our iterative IBEA.

View full abstract

The focus of this paper is mental tension detection in speech to assist control the tension in day-to-day business such as conferences and operations in a call center. It is difficult to use classical techniques for mental tension detection in day-to-day business because those techniques require invasion body by electrodes or squirts and tied up with cables. In order to achieve non-invasive, non-contact and low-restriction method, this proposed technique uses acoustic features in speech. The technique uses the vocal tract model which represents the shape and the tightness of throat muscle. It is required to devide multilevel tensions into some degrees. In this paper, multiscale tensions are labeled by physiological monitoring such as electrodermal activity and heart rate.

View full abstract

We propose a motion pattern modification with acquired sensor information. Living environments for human beings are complicated for human-friendly robots. The difference of conditions affects the realizability of the robot motion pattern to do some tasks. A motion designed for normal condition such as on the flat plain floor. Robots can work there obivously. They however can't work well in another condition. Our purpose is that robot autonomously finds an invariant on the motions for task accomplishment. Our method find another motion pattern by referring the sensor information which is obtained in normal condition.

View full abstract

Conventional optimization based on a simulation modeling divided into elements has been changing into the optimization considering the entire system due to the rapid progress of computers. Parameter studies by computer simulations are necessary and indispensable for multi-objective optimization problems of the entire system such as system design and scheduling problem. We can calculate a large amount of solutions in a practical time by using parallel calculation technology. However, it is difficult to use the acquired solutions effectively, because the solutions have multi-dimensionality on evaluation values. This study tries to develop ``Mining of Solutions" technique with visualization. As with necessity of data mining for large quantities of data accumulated by information technology, ``Mining of Solutions" technique for a huge amount of solutions for multi-objective optimization problems will become important. This study proposes a visualization method for acquired solutions which have multidimensional evaluation values. The proposed method enables us to grasp the distributed structure of solutions and clarify the relationship among clusters of solutions.

View full abstract

In this paper we show a method for applying a neighboring crossover to QDSEGA for controlling multiple agents. In QDSEGA, a control table for Q-learning is dynamically restructured by a genetic algorithm. In original QDSEGA, the control table is restructured using just a simple crossover. We have already proposed a neighboring crossover to enhance the performance of QDSEGA, though, when we apply our neighboring crossover directly to QDSEGA for controlling multiple agents, it does not give a better performance. In this paper, we combine a simple crossover and a neighboring crossover in order to enhance the performance of QDSEGA. Computer simulation results show that our method reduces the size of the control table by 50 percents compared to QDSEGA with simple crossover.

View full abstract

Recently, many kind of robots have been developed. Motion patterns are designed to specific architecture. When the architecture of the robots has transformed broken by such as joint failure or frame deformation, the robot can't express their prepared motion patterns. In this paper, we proposed adaptation method for transformed architecture using motion expression with remarkable postures. Remarkable postures obtained by vector quantization can describe whole motion patterns by interpolation. Using remarkable postures, the prepared motion patterns can adapt for transformed architecture. Remarkable postures are modified to adapt transformed architecture, so that failed motion patterns are modified appropriately to transformed architecture without the information on transformed architecture. We assumed the robot to be have broken joint. In this condition, experimental results show the adaptation ability for transformed architectures by proposed method without reverse dynamics methods.

View full abstract

Fuzzy rule-based classifiers have been one of the most widely used classifiers because of their highly interpretable structures and high classification ability. Several different measures of a rule in a fuzzy rule-based classifier as the measure of the interestingness have been proposed and used. Among them are confidence, support, gain, variance, chi-squared value, entropy gain, gini, laplace, lift, conviction. It is shown that the best rules according to any of these measures are Pareto-optimal rules with respect to confidence and support maximization. In this paper, we the examine the effects of designing classifiers from Pareto-optimal and near Pareto-optimal rules on the classifier's accuracy-complexity tradeoff curve.

View full abstract