Interactive Evolutionary Computation (IEC) is a useful method for Kansei designing because it does not require any quantitative functions for optimizing systems. IEC optimizes the system based on intuitive human evaluation. User of the system with the IEC needs to evaluate directly a lot of objects that are presented by that system. Consequently the user has a problem with being distressed by fatigue. So we propose a method of evaluation that is a simple pair comparison with tournament style. In the experimental result, it is declared that the proposal method possesses ability for optimizing systems as well as the conventional IEC method. Furthermore, we execute questionnaire survey about usability of the tournament evaluation by using two interactive evolutionary computing systems with image and sound. As a result, it is confirmed that the proposed method is a useful interface because it can reduce user's fatigue with relation to evaluation.

View full abstract

The rational choice theory made clear about the action that voters support parties. In that theory, however, voters are supposed to know each political position of all parties and they can calculate the political distance from them to all parties. In this paper, we assume that the voters cannot know each political position of all parties, but know only the political distance between the political output and their political position. We inspect that the voters' adaptive support controls the political output and they can lead to the best political position for them. In this simulation, two parties competing for votes each other. We show the availability of voters' adaptive support in the case that a party change suddenly its political position. As a result, we come to that voters can lead the political output to their favored position no matter that a party change suddenly its political position.

View full abstract

In this paper, we focus on multilevel multiobjective programming problems where multiple decision makers in a hierarchical organization have their own multiple objective functions, and propose an interactive decision making method to obtain the satisfactory solution which reflects not only the hierarchical relationships between multiple decision makers but also their own preferences for their objective functions from among the generalized lambda-extreme point set. The generalized lambda-extreme points are defined by using lambda-extreme points of each of the decision makers, which are the extended concept of Pareto optimal solutions in multiobjective programming problems. In order to obtain the satisfactory solution from among the generalized lambda-extreme point set, an interactive fuzzy decision making method based on the hyperplane method is proposed.

View full abstract

In information society, an agreement from many individuals needs an agreement process with efficiency and consistency. Group decision-making support model is a model that constructs a model of an individual, makes differences between individuals and supports an agreement process. In this paper, we propose the model of a group decision-making support model based on the structural modeling. Use of the model makes it possible to do a flexible and an effective modeling for group decision-making support model.

View full abstract

One of the authors, previously, presented the Strategic Task Sequencing method, shortened to the STS method, based on an instructional structure graph. Contrary to this, he presented an estimation method for instructional strategy by using the STS method. This estimation method, shortened to the ESTS method, can be realized by introducing membership functions that represent characteristics of instructional strategies. However, strictly speaking, this method is not unified by Fuzzy Theory. This paper applies fuzzy decision-making to the ESTS and discusses this new method, called the FSTS method. Finally, this paper discusses validity and availability of this method from both theoretical and practical view points.

View full abstract

In this paper we propose the exponential fuzzy numbers in Analytical Hierarchy Process (AHP), employing the fuzzy concept. The objective was determined a method of deriving the weights of criteria and their alternatives. Next, the influence of fuzzy is examined by using the membership function. Specifically we use the exponential fuzzy numbers for the pairwise comparison scale. Same experiments with AHP were conducted using the pairwise comparison scale and the new pairwise comparison scale of the method was examined.

View full abstract

The AHP method has been widely used in decision making. But in many cases, data matrices do not have enough consistency, i.e. the data lose its reliability. For the cases, we have proposed an AHP using fuzzy reciprocal matrix before. Then in this paper, we consider more about optimal weights pattern and a weight representation.

View full abstract

In this study, we focused on not only behavioral interaction but also communication which has not beeb treated in game theory research field. And we investigated how local communication caused by network structures affects human decision making, and we analyzed that according to types of network structures. As a result, we found that if edges of networks are few, each players' payoff tend to be similar. On the contrary, if edges of network are abounding, each players' payoff tend to be different and dispersed. We presumed that it is because useage of communication channel are different in each plater.

View full abstract

Kohonen's Self-organizing Map (SOM) and its application to clustering problems have been studied well in recent years. We have already proposed a two-stage self-organizing map algorithm what we call Two-stage SOM, which is effective in application to clustering problems. However, clustering performance of Two-stage SOM has not been examined thoroughly. In this paper, we propose the improved Two-stage SOM in order to achieve more stably good performance for various kinds of dataset. Then we carry out numerical experiments to evaluate clustering performance of Two-stage SOM by using artificial datasets and several kinds of dataset from UCI Machine Learning Database. Through the comparison with k-means method and several hierarchical clustering methods, it is confirmed that the improved Two-stage SOM can achieve stably good clustering performance.

View full abstract

The purpose of this research is to realize a self-organizing map of dynamical systems. In other words, the architecture to identify the dynamical systems using the observed time series data set, and to acquire a self-organizing map based on the degree of similarity between systems is developed. This architecture can be applied to various fields like the orbit generation of robot arms and electroencephalogram analyses, etc. In this research, we tried to realize the self-organizing map of the dynamical systems using mnSOM. However we can not use the conventional mnSOM since its algorithm is not considered the problem of hidden variables which are necessary for identification of dynamical systems. Then, we developed the algorithm in consideration of hidden variables, and verified the effectiveness of the new algorithm.

View full abstract

A robot has to autonomously acquire the environmental map by interaction with its environment to per form the self-localization and the behavior planning in unknown environment. The methods for acquiring the environmental map using the topology preserving neural network, such as a Self-Organizing Map (SOM), a Neural Gas (NG) and so on, have been proposed in the research topic for acquiring the autonomously environmental map. In these topology preserving neural networks, however, it is difficult to estimate the coordinates and direction of robot from information with multi-dimensional vector. In this presentation, we propose the acquiring the environmental map using SOM2, in which each neuron unit is extended to an SOM network module. Moreover the algorithm of the proposed method and the experimental results are shown.

View full abstract

This paper describes a group decision support system (a negotiation support system) based on users' Kansei using SOM and its application to automobile purchase for family. First, we evaluate Kansei scores of automobiles existed in real world using Kansei engineering. Second, we make a SOM map based on both the Kansei scores and group member's Kansei scores of their ideal automobiles for family use. If the member's ideal automobiles are away from each other on SOM map, the system show some compromise proposals to the member for negotiation. Where, the compromise proposals are made by small changes of member's Kansei scores. Finally, system gathers the member's ideal automobiles with small changed Kansei scores into a small area on SOM map, then decides an existing automobile nearest to the area as an negotiation result of members.

View full abstract

In this paper we propose applying tree structured SOM(Self-Organizing Maps) for rule extraction from clinical data of bladder cancer patients. Bladder cancer is a cancer that has the highest mortality rate following prostate cancer in the field of urology. Because a tumor marker like PSA(Prostate Specific Antigen) for prostate cancer does not exist, an effective method for prediction has not been established. We aim at analyzing and extracting rules from clinical data of bladder cancer patients. The final goal of the research is to establish a reliable prediction method. It is difficult to apply previous decision tree generation methods such as ID3 to clinical data of the bladder cancer patients used in this research because the clinical data consists of properties with both continuous values and discrete values. In this paper we propose a method that trains multiple SOM arranged in a tree structure, from which the decision tree automatically generated. We investigate the effectiveness of the proposed method by applying to rule extraction from clinical data of the bladder cancer patients.

View full abstract

At present many environmental problems such as global warming, ozone depletion and acid rain occur on a global scale. Using satellites to remotely observe the earth is one of the major efforts to monitor and research the causes of global environmental change. However, there are still many problems remaining. One area with room for improvement is precision of land cover classification from remote sensing data. In this research, we use self-organizing map (SOM) to analyze land cover classification or land use change from MODerate resolution Imaging Spectroradiometer (MODIS) data. The amount of carbon dioxide held in vegetation differs for types of vegetation, and the goal of the research is to improve classification accuracy in order to predict the total amount of carbon sink on land cover.

View full abstract

To a localization problem of an active RFID tag in the indoor environment we approach using SOM (self-organizing map) in the following framework. We deploy several anchoring tags (whose locations are known) in the room. A receiver is set on a robot. The following process is repeated until the robot reaches the destination: 1. We measure RSSI (Received Signal Strength Indicator) values for all tags. 2. We make a clustering of tags by SOM. 3. According to this map we decide the direction of the destination. 4. The robot (equipped with a receiver) moves toward the destination.

View full abstract

Recently many studies have been made on the associative memory. However, we were beginning to feel the limits of the performance of the conventional associative memory that is typified by Hopfield's model. Ritter proposed a morphological associative memory. In his research, he presented associative memories W and M. The associative memory W is robust under erosive noise and M is robust under dilative noise. However, these memories are not robust under general noise that includes both erosive and dilative noises. This conventional binary morphological associative memory was made associative memory with crisp mini-max algebra operations. This paper presents a new approach to morphological associative memory based on fuzzy set theory. We show the validity of this method by numerical simulations. This method is expected to show the high percentage of recalling under general noise.

View full abstract

It is known that an introduction of small-world network to the morphological associative memory reduces the memory matrices and improves the perfect recall rate. However, in previous studies, the degree of the presence of small-world network has not been discussed. In this paper, we measure the degree of the presence with L (characteristic path length) and C (clustering coefficient) and investigate the relation between these parameters (L and C) and the performance (the perfect recall rate).

View full abstract

In this paper, we propose a new method of the three-level halftoning based on error diffusion method. In the proposed method, two nonlinear conversion images are generated from an original image, at first. Then, those continuous-tone images are converted to bi-levels using error diffusion method, and the resultant image represented only by three levels is obtained by averaging the binarized images. The effectiveness of the proposed method is verified by applying it to gradational and natural images.

View full abstract

Transcranial ultrasonography can non invasively image the brain in real time processing. However, the ultrasonic wave causes refraction in the skull, so the conventional ultrasonography can not image the brain from arbitrary place. In this paper, we propose an imaging system by considering ultrasonic refraction. In this experiment, we use a cow scapula to imitate the skull and a biological phantom to imitate cerebral sulcus. To calculate the ultrasonic refraction, we require the scapula thickness. We can easily determine the echo from scapula surface. However, it is difficult to determine the echo from scapula bottom. In this paper, we calculate the scapula thickness by using fuzzy inference. In the inference, we employ amplitude, correlation coefficient and the elapsed time. As the result, our method can estimate the thickness of scapula at all point and can essfully visualize the phantom surface image from arbitrary places.

View full abstract

We estimated the cardiac risk in non-cardiac surgery. We developed predictors for perioperative cardiac accidents, including hard events (cardiac death and myocardial infarction), using a total of 22 clinical properties such as patient's clinical information, surgical risk, and results from nuclear scanning. A total of 1351 surgery records from Keio University Hospital, Japan were used and we analyzed them using linear and support vector machine (SVM) classification models. The input parameters of these classifiers were selected with the incremental stepwise method. Both linear and SVM classifier achieved similar prediction results of 80% in sensitivity (SE) and 66% in specificity (SP) for all cardiac accidents and 85% in SE and 81% in SP for hard events with a leave-one-out cross-validation strategy. Although the selected parameters by each classifier were slightly different, several parameters measured by nuclear scanning were commonly selected and we can therefore conclude that the nuclear scanning is effective for estimating perioperative cardiac risk in advance.

View full abstract

To diagnose the cerebral diseases using MR images, skull stripping is one of the fundamental image processing techniques. There are many conventional skull stripping methods for adult brain MR images. However, image features of infantile MR images are different from those of adult MR images, and in case of infant, the skull is close to the cerebral surface. Thus, these conventional methods for the adult brain cannot be applied to the infantile brains. In this study, we propose an automated method for stripping the skull region from infantile MR images. The proposed method is based on fuzzy rule-based active surface model (FASM) that has an ability of representing the physicians' knowledge with fuzzy IF-THEN rules. The experimental results of 8 infantile MR images validated that the proposed method extracted the cerebral region with a sensitivity of 98.84% and false-positive rate of 1.21%.

View full abstract

Interactive Evolutionary Computation can search a target in non-linear and multimodality spaces such as human subjectivity and KANSEI. In our study, we focus gamma correction as image adjustment and apply IEC to the image quality improvement support system which can derive the image quality adjustment parameters, gamma value, on the basis of human subjectivity.

View full abstract

An adaptive particle filter with a statistical motion model, called the optical flow--driven motion model, is proposed for sequential Bayesian visual tracking. The motion model predicts the current state with the help of optical flows, i.e., it explores the state space with information based on the current and previous images in an image sequence. This exploration improves the prediction accuracy, compared to a prior model based exploration, which is widely used in visual tracking. In addition, an automatic method for adjusting the variance of the motion model is introduced. The variance affects the performance of tracking but the parameter is manually determined in most particle filters. In experiments with two real image sequences, the proposed motion model is compared with a random walk model. The experimental results show the proposed model outperform the random walk model in terms of accuracy even though their execution times are almost the same.

View full abstract

In these years, a composite of artificial culture bone and bone marrow stromal cells (BMSCs) are hoped to effective treatment to large bone defect as impossible to curing for natural healing. Conventional evaluation method for the cellular quantity is firstly to crush the artificial culture bone injected BMSCs. So, the approximate amount of cell for curing is unknown. Therefore, this paper describes an ultrasound evaluation system for cellular quantity of artificial culture bone. Our system firstly measures ultrasound wave into the composite. And, we obtain two characteristics values. One of these values, the amplitude value, is directly measured by obtained wave. Another the frequency value is calculated by frequency spectrum converted with cross-spectrum method. We employ fuzzy inference using correlation of these values and true quantity for the cellular quantity. As a result, our method is able to evaluate the quantity with high accuracy.

View full abstract

In this paper, a method which specifies the signboard region and extracts the characters inside the signboard is proposed. We usually take notes not to forget what we should leave to memory. But it is often that the work is too intrusive. Our aim is development of a new input-interface so as to input texts from a picture. Most of signboards are composed of almost monochromatic region. On the basis of the observation, image segmentation using color information is applied, and then we get some binary images by applying threshold for each segmented region. Each binary image is enclosed by the smallest circumscribed square. The signboard region is specified according to distribution and area of the white pixels inside the square. As a result of experiment, we confirmed the effectiveness of the proposed method.

View full abstract

2-D/3-D image registration is a technique to estimate 3-D pose/position of 3-D model from 2-D image. It has been used in various fields including industrial and medical fields; especially, it has been applied for analyzing knee joint kinematics. Roughly, matching score used in 2-D/3-D image registration can be classified into two approaches: landmark based method and intensity based method. In case of landmark based method, the matching score is defined as a distance between landmarks in the images. Although the method is a simple and high speed processing, the registration accuracy is depend on extraction accuracy of landmarks. In contrast, in case of intensity based method, the matching score is defined as differences of intensity between the images for all pixels. The method has an advantage that it is unnecessary to extract any landmarks, however, it has a tendency toward the falling into local solutions. In this paper, we introduce attended/unattended ROI (region of interests) into matching score. By representing the ROIs with fuzzy spatial maps, we can decrease the dependency of registration accuracy on defining ROIs. In addition, by employing ROIs, the proposed method improves the registration accuracy in comparison with intensity-based method. To validate the performance of the proposed method, it has been applied to image matching between 3-D knee bone model reconstructed from multidetector-row CT and 2-D digital radiographic image. The experimental results showed that the proposed method estimated knee angles with a root mean squared error (RMSE) of 0.4 deg that was superior to conventional intensity-based method.

View full abstract

In a digital still camera (DSC) system, in most cases, a dynamic range is compressed linearly in logarithmic space. Therefore, an image taken by DSC tends to be low contrast, and gives us a flat impression without liveliness in comparison with silver halide photography. In this study, we propose a new digital image enhancement method by using sigmoid functions. In the proposed method, parameters of sigmoid functions are decided by local information of the input image for each pixel. Further, multiscale image enhancement is carried out by setting some scales of local regions, and the resultant image, that is visually good and high contrast, is provided by synthesizing enhanced images obtained by changing scale.

View full abstract

It is pretty important to measure vital-sign continuously for keeping humans good health. There have been developed many equipments for vital-sign measurements. Among vital-signs, heart rates while asleep are focused on here because they will bring some information of humans' health condition. In this paper, an unconscious and non-invasive measurement technology is discussed to develop the equipment of heart rate measurement using simple and low-cost air pressure sensors. According to the experimental results, the technology is proven as to provide enough accuracy.

View full abstract

Various computational intelligences are formulated based on max-plus algebra, especially, neural networks, wavelets, cellular automata including fractal/L-system/petri-nets, are formulated based on max-plus algebra. The advantages of ordered structure based computational intelligences are mainly correspond to high speed/parallel processing, analyzing information in terms of ordered structure, and treating only discrete information (no quantization error). Through application experiments, it is shown the potential of the ordered structure based computational intelligences.

View full abstract

Middle managers exhausted attending many meetings. Effective management for these meetings is required utilized by Information Technology. Up till now, although many meeting support systems have been proposed, they support simple functions such as video camera systems, electronic shared whiteboard and so on. Participants can share the on going information during the meeting. Much effective supports will result the meeting fruitful. We modeled the meeting process into two phases. One is divergence phase that all participants propose a lot of ideas including brain storming and developing them. The other is convergence phase that they try to made summaries. In this paper we clarified the problems in each phase and investigate them. Some issues are reported arose from these investigations.

View full abstract

Traffic congestion causes terrible social and economic costs on our society. To reduce traffic congestion, several approaches have already been used. Among them, Vehicle Information and Communication System (VICS), which provides traffic information such as congestions and restrictions, is now widely and successfully used in Japan. VICS gathers traffic information for a traffic information center, processes the gathered information, and transmits it toward vehicles. Two major difficulties about VICS, however, are reported. One is the necessity of a huge amount of investment on infrastructures such as a traffic information center. The other is a time lag between the observation of road conditions by VICS and their utilization by each driver. To resolve these difficulties, several studies discussed the potential ability of wireless communication between vehicles, usually referred to as Inter-Vehicle Communication (IVC). In this paper, we propose a route guidance method based on traffic information sharing through IVC. Experimental results show the effectiveness of the proposed method.

View full abstract

Location information obtained from GPS(Global Positioning System)on mobile phone have great potential to enhance our daily life. One popular application using these information is a walking navigation system. It is pretty easy to find the destination with the systems, even when we visited there for the first time. On the other hand, when we are just enjoying walking around, the system is not so helpful. In this paper, we have investigated the functions we need to support where we were enjoying town walks. Some answers from questionnaires suggest that simple information and operations are required. The functionality to be implemented are discussed.

View full abstract

The fuzzy traffic simulator would be created for considering the specific traffic safety measures to the intersection. Traffic accidents can be realistically observed in the virtual intersection as results of fuzzy traffic simulation. The mechanism of traffic accident would be analyzed to find the essential factors of traffic safety. Therefore, efficient traffic safety countermeasures can be summarized from the observation of virtual accidents in the fuzzy traffic simulation with different initial conditions for behaviour of vehicles.

View full abstract

As a heat load prediction method in district heating and cooling systems, the efficiency of layered neural networks has been shown, but there is a drawback that its prediction becomes less accurate in periods when the heat load is nonstationary. In this paper, we propose a new heat load prediction method superior to the existing method based on a layered neural network by using a recurrent neural network to deal with the dynamic variation of heat load as well as new input data in consideration of characteristics of heat load data.

View full abstract

We have many kinds of data of time series such as stock prices. We don't understand time series with a stochastics process model. We have proposed a method to express a global trend of time series in a natural language. In the previous research we expressed global trends via a representive value on the fuzzy intervals in the temporal axis. We express local features for the position of large difference between the original data and the data generated from the global trend. We can't express data of oscillation well with them. We propose a method to express the scale of amplitude with the standard deviation of difference between the original data and the data generated from the global trend and the degree of oscillation with the standard deviation and the numbers of oscillation. We express data of time series with the global trend, local features and oscillation.

View full abstract

Decision trees are one of the most popular methods for acquiring knowledge from data. In this paper, we try to acquire knowledge from data of time series using a global trend and local features and oscillation of time series in addition to the current value in each attribute. This method generate rules such as "if a global trend is slightly increasing and there is a big wave in the last stage, then this data is classified into C1". We use a method based on Fuzzy ID3 algorithm and the expression of the data of time series in a natural language in the previous research. We apply the method to the data of stock prices, where we acquire rules from the data of attributes, global trend, local features, oscillation and category of business, and the class.

View full abstract

In this paper, sales of every day of a certain apparel store were investigated for the consumer's demand forecasting model construction, and the meteorological factor such as the temperature was investigated at the same time. And, consumer's demand forecasting model is constructed by using meteorological factor in statistically clarifying the relation between these. Concretely, the fuzzy reasoning model to whom demand is forecast by trying the analysis of sales of every day and the relation between the number of visitors and the meteorological factor by applying a decentralized analysis, selecting the meteorological factor from this analytical result, and using the obtained factor is constructed.

View full abstract

Rough set theory has mainly been applied to data with categorical values. In order to handle data with numerical values, we define numerical patterns with two symbols # and @, and propose more flexible rough sets based rule generation. The concepts of coarse and fine for rules are explicitly defined according to numerical patterns. This paper focuses on the rough sets based method for rule generation, which is enhanced by numerical patterns, and refers to the tool programs. Tool programs are applied to data in UCI Machine Learning Repository, and some useful rules are obtained.

View full abstract

Rough sets are applied to data tables containing missing values. Discernibility and indiscernibility of missing values are considered simultaneously. A family of possible equivalence classes is obtained, in which each equivalence class has the possibility that it is an actual one. By using the family of possible equivalence classes, we can derive lower and upper approximations, even if the approximations are not obtained by previous methods. Furthermore, the lower and upper approximations coincide with ones obtained from methods of possible worlds. Keywords: Rough sets, Missing values, Possible equivalence classes, Lower and upper approximations.

View full abstract

Rough set theory was proposed by Z.Pawlak in 1982. This theory can mine knowledge as a decision rule from a database, a web base, a set and so on. The decision rule is used for data analysis as well. And we can reason for an unknown object using the decision rule. The objective of this paper is to apply the rough set theory to analysis of time series data. It is possible to acquire knowledge from time series data using regression line and apply a method to predictions.

View full abstract

We introduce variable precision rough set models to a recommendation system using Pawlak's rough set theory we have proposed. By extracting reducts and decision rules based on user's queries, the proposed recommendation system recommends some goods that the user may like though the characters are different with the queries. However, too many goods may be recommended, and some of recommended goods have low connection with user's queries. To avoid these problems, we propose a new recommendation method using variable precision rough set models, and evaluate the proposed method by experiments.

View full abstract

This paper discusses the selection support system by which everyone can select leather shoes easily suitable for events. The purpose of this research is to clarify an uncertain image of leather shoes in the fashion said that the total balance is important. Each combination of attribute value of suitable leather shoes for the image of the events and unsuitable leather shoes is displayed, and in the system that proposes it, everyone doesn't hesitate and be able to put on suitable them for the events by choosing them in consideration of the annexation rule. It was guessed that this system was able to be constructed if the image of the event was first investigated, and to discover leather shoes corresponding to the image. And, the rough set was used so that the image of event connect leather shoes.

View full abstract

In an English class, the students learn various sentence structures. In order to be able to use the sentences freely, they have to make an effort to memorize the sentences repeatedly. It is said that e-Learning has a merit that the learner can study anytime and anywhere again and again. Then, we tried to make a construction of e-Learning system for supporting the self-study of memorizing English sentences. Besides, we developed the system that we could examine the results of their learning using the EXCEL(R).

View full abstract

This article presents the data analysis of a consciousness survey where university creation in view of humor is asked to the subject. We obtained, from the fator analysis, that (visual) design, correspondence to the out-university and so on could be a keyword for this problem solving.

View full abstract

Recently, e-Learning system, which is not depended upon place and time for learning, commonly has been used as a great learning method. e-Learning means taking lessons on the computer through the networks, and it can be generally classified into an asynchronous type and a synchronous type. The asynchronous type mainly uses documents and the synchronous type does images and the movies for lessons. Previously, authors proposed a method for taking the appropriate images of written area on the blackboard. In the method, the blackboard is segmentalized into several areas and template matching method is used to acquire the high-resolution images of the written area. It is however, difficult to acquire appropriate rules or search areas for template matching method. This makes decision of template size and matching areas more difficult with high accuracy. This paper, propose the new method to decide the desired templates and matching areas automatically by using Evolutionary Computation. The experiments were conducted to confirm the effectiveness of the proposed method and created templates are discussed.

View full abstract

We have proposed an intelligent intrusion detection system (IIDS) that used soft computing approaches. The current IIDS uses the support vector machine (SVM) or the CLustering In QUEst (CLIQUE). To evaluate IIDS, we need to use real world network traffic data because each IIDS is used at various environments. That is, it is very important that we use the real world network traffic data for training and evaluating. Therefore, we make software that aids us to generate training patterns from real world network traffic. Then our IIDS is trained by using of the training patterns. In computer simulations, we show that when the real world network traffic data were used for training, the classification performance is better than when the DARPA data were used.

View full abstract

This paper proposes an implementation method of dribble skills for RoboCup soccer simulation. The developed dribble skills consist of two actions: One action is to move forward and the other is to adjust the body angle of agents. The domain knowledge of RoboCup soccer simulation is used for the moving forward action while neural networks are used for adjusting the angle of dribbling agents. Training data for the neural networks are obtained from game logs of a target team. We perform computational experiments to examine the effectiveness of the implemented dribble skills.

View full abstract

We proposed a method to extract knowledge about associations between attributes of data by fuzzy sets and fuzzy qualifiers, where it is expressed in a form of natural language for a human to understand easily the meaning about the data. The knowledge associates local values in some attribute with those in another attribute. In this paper, we extend the method by extracting global associations between two attributes. For example, "The data of high degree of property that the greater value in the attribute A gets greater in B is classified into the class C1". We normalize values for each attribute so that the average is 0 and the standard deviation is 1. We express normalized data in polar coordinates to extract associations between attributes by restricting arguments with fuzzy sets. The method extracts knowledge about global associations between attributes.

View full abstract

Recent study has proposed the ADG ( Automatically Defined Groups ) method on the basis of the GP ( Genetic Programming ) for extracting diagnostic rules from computer log files. However, because the study has dealt with each word in log files as a GP terminal, the number of terminal has been too large to obtain the evolutionary optimization sufficiently. Therefore, we propose a word-selection-method by focusing on the similarity score of words in order to solve this optimization-problem in this study. Moreover, we discuss our selection method by comparison to other well-known scores in information retrieval, for example, TF-IDF ( Term Frequency - Inverted Document Frequency ), particularly in terms of the optimization. According to this purpose, we performed our experiment for extracting rules by using the data output from several servers. As the result of this experiment, we confirmed that our similarity score was superior to other ones.

View full abstract

Recently, the computer system is composed of two or more computers including network equipment, and provides one total service by their complex operations. However, each computer is equipped with various hardware parts such as CPU, memory, and HDDs. If one of parts breaks down, a computer will not operate. Moreover, it is difficult to monitor the normal state of many system operations because various Operating Systems(OSs) are worked in their computers. In this paper, we propose a web based management tool of working computers with Open Source Software to detect the hardware fault and diagnostic rules extracted by Automatically Defined Groups to detect software errors.

View full abstract