Information recommendation systems are used to provide suitable information to users. Conventional systems entail the difficulty that once an item is selected, only similar items are recommended. This tendency leads to user “boredom” with the recommendations and markedly reduces the recommendation effectiveness. Therefore, the present study focuses on “serendipity”, an indicator such as unexpectedness in information recommendation, to eliminate user boredom. Novelty and diversity are used to evaluate serendipity, and the unique preferences of the target user, which have not been considered in previous research, are taken into account. The conventional method, existing method from previous studies and six proposed methods are then compared. The findings indicate that the best results were obtained by recommending items that other users did not like but which had a high predicted evaluation value. This method resulted in all recommended items having novelty, and the diversity was also improved by about 14% compared to the conventional method, and serendipity was improved the most. Furthermore, it maintained the same recommendation accuracy as the conventional method. This method is expected to improve user satisfaction.

Detection of seismic radiation emitted from the earth's crust can be used for predicting earthquakes. We have been observing electromagnetic (EM) waves in the extremely low frequency (ELF) band. However, the database of seismic radiation signals is not large enough for developing detection methods for anomalous signals from the EM wave data. Previous studies on anomaly detection based on hidden Markov model yielded good results but have not addressed certain problems with the input data. In this paper, we discuss these problems and propose a revised method which achieves better performance in detecting anomalous signals from the waveform images of the ELF magnetic signals.

Previously, we developed two NMF algorithms named QL1-NMF using quasi-L1 norm for analyzing environmental ELF magnetic field measurements. When the data includes many outliers, the QL1-NMF algorithms returned better results than other BSS algorithms using L1 norm. However, the derivative of cost function in our first algorithm was not based on a monotonically increasing function. This problem decreased the validity of algorithm. Our second algorithm had problems about stability though it was based on a monotonically increasing derivative. The method therefore required improvement of validity and stability. In the work described in this paper, we newly introduced the update functions that were based on a monotonically increasing derivative. Computer simulation results and real data's results confirm the new algorithm works more stable than the previous one.

Recently, as a robust backup, geographically dispersed and distributed backup systems to save data have been attracting attention. However, existing distributed backup systems that can be operated at low cost present various difficulties such as processing time, immediacy, and storage capacity. Therefore, as an existing method intended to solve this problem, multiagentbased distributed backup systems exist for individuals to solve problems of processing time and immediacy. However, problems of storage capacity exist in such systems. This research presents a deduplication technique of block data to solve problems of storage capacity. By incorporating this into the system, we aim at storage capacity reduction of backup data.

In recent years, programs operating on networks are increasing because of the development of distributed processing technology. Therefore, agent-oriented programming, which can respond dynamically to environmental changes, is receiving attention as a development technique. The agent framework is supports software development through agent-oriented programming. For this research, DASH is a repository type agent framework. A salient feature of DASH is its mechanism, called a repository, which accumulates agents. Moreover, it has the ability to constitute an agent system dynamically using the accumulated agents. Nevertheless, the conventional DASH has a persistent shortcoming. Accumulation of agents who covers all agents is not performed because the agent which each developer developed is accumulated only in a developer's repository. To solve this problem, information is acquired and unified from each repository that exists in a distributed environment. Consequently, we proposes a system that supports the practical use of an agent existing in a broader-based distribution environment. In the operation experiment, we confirm the operation of the search function and the single call function of the agent and increase in the number of available agents by the proposed mechanism.

The demand for backup in preparation for data disappearance caused by a large-scale disaster, etc. is increasing greatly. For robust backup, geographically distributed backup to save data is indispensable. If the backup is distributed, then total loss of data after a catastrophe is unlikely. For this study, we assume that a user is a college student, and assume distributed backup among plural universities. Although such a user distribution backup is applicable at low cost, it entails some difficulties. We propose to resolve these problems using agent technology.

We developed a scenario editor to improve the environment of scenario editing for a spoken dialogue system. The purpose of this study is to improve the editing environment of the spoken dialogue system. We have developed MMDAgent (a fully open-source toolkit for voice interaction systems), which runs on a variety of platforms such as personal computers (running Windows, Mac OS, or Linux) and smartphones (running Android). So that MMDAgent can be used in a similar environment, the scenario editor is also implemented on a Web browser. Experiments were conducted for subjects using the scenario editor of our system. It was found that our system provides better readability of a scenario and allows easier editing.

In this paper, we propose a voice interaction system using 3D-CG virtual agents for stand-alone smartphones. Because the proposed system can handle speech recognition and speech synthesis on a stand-alone smartphone differently from the existing mobile voice interaction systems, this system enables us to talk naturally without encountering delays caused by network communications. Moreover, proposed system can be fully customized by dialogue scripts, Java-based plugins, and Android APIs. Therefore, developers can make original voice interaction systems for smartphones easily based on proposed system. We have made a subset of the proposed system available as open-source software. We expect that this system will contribute to studies of human-agent interaction using smartphones.

Recently, information recorded on paper has been changed rapidly to digital data. Backup systems emphasize prevention of data loss by disasters, mistakes of user operations, and so on. For robust backup, it is necessary that the data be distributed geographically. Using a distributed backup system, the possibility of all data becoming lost will definitely be low. However, several problems with distributed backup systems remain. First, preparing multiple computers is expensive. Second, the computational cost of encryption and redundancy is high. Third, user workloads necessary to manage and maintain the system are great. As described herein, to overcome those problems, we propose a distributed backup system using multiagent technology. The agents can work autonomously depending on the environment.

The agent-oriented computing is a technique for generating the agent who operates autonomously according to the behavior knowledge. Moreover, agent can have the characteristic called "Learning” skill. More efficient operation of agents can be expected by realizing "Learning" skill. In this research, our aim is to support agent designer who designs and develops the intelligent agent system equipped with “Learning” skill. We propose interactive design environment for agent system with learning mechanism using repository-based agent framework called DASH framework. Proposed framework enables agent designer to design and implement the learning agents without highly expertise, therefore we can reduce the designer's burden. In this paper, we explain the DASH framework, Q-learning, Profit Sharing and proposed design environment. Moreover we show the effectiveness of the proposal method through the some experiments.

In this study, the authors forecast students’ future academic records using past attendance recording data and grade data. We use a Bayesian network as forecasting method. During construction of the Bayesian network forecasting model, unnecessary variables become noise and so lower the forecasting accuracy. Therefore, to improve the forecasting accuracy, we used information gain to reduce the number of variables in the model. As a result, accuracy improved.

Detection of an anomalous signal radiated from the earth's crust is useful for predicting the precursor of the earthquakes. Using the Extremely Low Frequency (ELF) band, we have observed the electromagnetic (EM) wave. Various methods for detection of an anomalous signal have been proposed. The known problems for those techniques are related to a false detection due to the limitation of training data for great earthquake. We proposed the HMM based anomalous signal detection whose training data are the amplitude density distribution extracted from normal EM wave signals. However, acceptance probability in this technique is not corresponded appropriately to a change of anomalous signal. In this paper, anomalous signal detection using amplitude density distribution calculated by short term signals to track a temporal change is proposed.

It is known that the detection of an anomalous electromagnetic (EM) wave radiation from the earth's crust is useful for predicting the precursor of the earthquakes. We observed the EM wave using the Extremely Low Frequency (ELF) band. Various methods for detection of an anomalous signal have been proposed. Those methods have some problems. We proposed the anomalous signal detection based on HMM using the amplitude density distribution of an EM wave signal, which does not contain an anomalous signal. The effective ness of HMM approach for detecting an anomalous signal is already reported. However, the relationship between an anomalous signal and an output of HMM is not considered enough. In this paper, the temporal change of acceptance probability for anomalous signals is represented to show the potential of HMM approach for a temporal anomalous signal detection.

Anomalous environmental electromagnetic (EM) radiation waves have been reported as the portents of earthquakes. We have been measuring the Extremely Low Frequency (ELF) range all over Japan. Our goal is to predict earthquakes using EM radiation waves. The recorded data often contain signals unrelated to earthquakes. These signals, as noise, confound earthquake prediction efforts. It is necessary to eliminate noises from observed signals in a preprocessing step. In previous researches, we used ISRA, an algorithm of the Non-negative Matrix Factorization (NMF), to estimate source signal. However, ISRA is not robust for outliers because ISRA's cost function is based on square distance. In order to improve robustness, we should use lower order cost function. In this paper, we propose a nonnegative matrix factorization method using quasi-L1 norm in cost function (quasi-L1 NMF). In the experiment using ELF observed signals that include outliers, the proposed method extracts source signals more accurately than ISRA.

A seismic radiation emitted from the earth's crust is useful for predicting earthquakes. The electromagnetic (EM) wave in the Extremely Low Frequency (ELF) band have been observed at many places in Japan. In this paper, we propose the anomalous signal detection method based on HMM whose input vector is the amplitude density distribution of an EM wave. The amplitude density distribution is calculated from the image of an EM wave data. The optimal scale of an image to calculate an amplitude density distribution and the optimal number of states of HMM are investigated to achieve an accurate detection.

A novel class of bit-flipping (BF) algorithm for decoding low-density parity-check (LDPC) codes is presented. The proposed algorithms, which are referred to as gradient descent bit flipping (GDBF) algorithms, can be regarded as simplified gradient descent algorithms. The proposed algorithms exhibit better decoding performance than known BF algorithms, such as the weighted BF algorithm or the modified weighted BF algorithm for several LDPC codes.

Anomalous radiation of environmental electromagnetic (EM) wave is reported as a portent of earthquake. We are observing environmental electromagnetic waves in ELF band all over Japan. Our goal is to predict earthquakes by detecting anomalous EM radiation. We have proposed various detection methods of anomalous radiation. However, the earthquake prediction requires the estimate of EM radiation area in addition to the detection of anomalous radiation. If the EM radiation area can be estimated, we can obtain information about the occurrence area of possible future earthquake. In this paper, we propose the method of detection and area estimation in the anomalous EM radiation by using cross-correlation in magnetic field azimuth. We apply the proposed method to the observed signals before the large earthquake that occurred in the past, and we detect an anomalous radiation and estimated the EM radiation area.

Decoding performance of linear programming (LP) decoding is closely related to geometrical properties of a fundamental polytope: fractional distance, pseudo codeword, etc. In this paper, an idea of the cutting-plane method is employed to improve the fractional distance of a given binary parity-check matrix. The fractional distance is the minimum weight (with respect to lscr1-distance) of nonzero vertices of the fundamental polytope. The cutting polytope is defined based on redundant rows of the parity-check matrix. The redundant rows are codewords of the dual code not yet appearing as rows in the parity-check matrix. The cutting polytope plays a key role to eliminate unnecessary fractional vertices in the fundamental polytope. We propose a greedy algorithm and its efficient implementation based on the cutting-plane method. It has been confirmed that the fractional distance of some parity-check matrices are actually improved by using the algorithm.

Anomalous environmental electromagnetic (EM) radiation waves have been reported as the portents of earthquakes. We have been measuring the Extremely Low Frequency (ELF) range all over Japan. Our goal is to predict earthquakes using EM radiation waves. Previously, we proposed a method of detecting anomalous signals by focusing on linear prediction errors. However, this method also sensitively responds to earthquake-unrelated anomaly. For accurate earthquake-prediction, we should eliminate earthquake-unrelated signals. In this paper, we try to reduce false detection rate by global signal elimination using Non-negative Matrix Factorization (NMF) and evaluate the effectiveness of this method.

Anomalous radiation of environmental electromagnetic waves is reported as a portent of earthquakes. Then, we are observing environmental electromagnetic waves of the ELF range all over Japan. We have been regarding the high amplitude in the observed signal as anomalous radiation. However, the observed signal contains a lot of low amplitude anomalies hidden by background noise. These anomalies also have the possibility with significant information of a portent of an earthquake. In this paper, we propose the method of detecting anomalous radiation using statistical property in magnetic field azimuth which doesn't depend on the amplitude level. Moreover, we show that an anomalous radiation is the portent of earthquakes by calculating an azimuthal cross-correlation among two or more observation sites.

The Japanese Islands are located in an active position of seismic activity and have therefore often been afflicted by the damage of an earthquake. There is a report that disorder of an electromagnetic (EM) wave occurs in nature as a precursor of crustal activity. In order to detect the sign, we have measured EM waves of 223 Hz at about 40 sites in Japan. Our final goal is to sense seismic activity and to predict earthquakes. To locate local sources radiating anomalous EM waves, we need to separate global noise components due to tropical thunder in the equatorial region from observed data. Therefore, in this paper, in order to separate global noise components, we perform independent component analysis. Our goal in this paper is to estimate a local anomalous EM wave source.

Occurrence of anomalous Electro-Magnetic (EM) radiation is known as a precursor of earthquakes. Our goal is to locate anomalous EM source caused by seismic activity for earthquake prediction. For the purpose, we have been measuring EM wave of 223 Hz at about 40 sites in Japan. Measured signal contains much noise caused by far thunder lightning around equators, human activities, near field lightning and so on. This paper shows that the principal component analysis methods for covariance matrices of measured signals are effective for removing a component caused by thunder lightning around equator and for separating each EM source. Moreover, it is shown that a result of elimination of noise caused by far lightning around equators, estimation of thunder lightning location in near field agrees with real thundercloud observed by radar system, and estimation of sources of anomalous EM radiation occurred before earthquake.

The tectonic activities that precede significant earthquakes release electromagnetic (EM) waves that can be used as earthquake precursors. We have been observing EM radiation in the extremely low frequency band at over 40 observation stations in Japan for predicting significant earthquakes. However, the recorded signals contain several noise components generated from the ionosphere, human activity, and so on. Most background noise in observed signal is attributed to lightning in the tropics. In this paper, we introduce a state space model and Kalman filter to reduce the noise. The good performance of the proposed model is confirmed.

The Japanese Islands are located in an active position of seismic activity and therefore often have been afflicted by the damage of an earthquake. There is a report that disorder of an electro-magnetic (EM) wave occurs in nature as a precursor of crustal activity. In order to detect the sign, we have been measuring EM waves of 223 Hz at about 40 sites in Japan. Our final goal is to determine the EM source more correctly. To locate the radiation source by multi point observations of EM power, EM wave propagation characteristic needs to be clear. In the presumed method of a generation source, we assume the attenuation model that EM wave intensity declines by the /spl alpha/ power of the distance from the EM source. /spl alpha/ is an unknown value, however, this paper analyzes /spl alpha/ by locating the thundercloud position which generates the EM wave.

Our goal is to locate anomalous electromagnetic (EM) sources caused by seismic activity for earthquake prediction. For the purpose, we have measured the EM wave of 223Hz at 35 sites in Japan. The measured signal contains much noise caused by lightning around the equator, human activity, near field lightning and so on. EM noises from far sources can be eliminated by using correlation in multi-point observation. But near-field noise is not easy to eliminate. This paper shows that the principal component analysis methods for covariance matrices of measured signals are effective for clustering signals and separating each source. It is shown that a result to estimate thunder lightning location in near field agrees with real thundercloud observed by radar system. And results of estimation of sources of anomalous EM radiation occurred before and after earthquake are also shown.

Our goal is to locate anomalous electromagnetic (EM) source caused by seismic activity for earthquake prediction. For the purpose, we have been measured EM wave of 223 Hz at 35 sites in Japan. This paper analyses relation between EM power attenuation and distance from its source. The relation is necessary to estimate the source location.