Recently, there have been many researches of robot partner for social communication. The robots should gather various type of information for communicating with people by using some internal sensors. However, it is difficult for the robots to gather the information by themselves because of the frame problem. Therefore, a system integrating sensor networks and portable sensing devices is required. The system should also have a platform for extracting and estimating the required information from the gathered data. In this paper, we applied a learning method based on the time series of data measured by sensor networks for estimating human behaviors. In the method, we propose a hierarchical learning structure based on spiking neural networks for modeling the human behavior patterns.
In the case that a robot controller is trained by means of evolutionary computing, the robot will be able to act sufficiently in the environment where the robot has been trained. However, if the robot is put in the environment which is more complex than the training environment, it cannot act sufficiently and is required to be trained again so as to fit to the complex environment. Based on this fact, we build a training environment for a robot controller with the partial components of the complex environment and aim to obtain a controller which makes a robot be able to act in the complex environment by only training the controller at the training environment. We clarify a way of building a training environment which functions effectively for training a robot controller and discuss how much training is necessary in the training environment to act under a more complex environment.
The honey bees that are a kind of social insects are well known to appropriately allocate their work force to some nectar sources by “waggle dance”. In addition, they have the threshold of the usage about the nectar sources and they can regulate it autonomously as the whole of bees. Mimicking this threshold regulation in honey bee foraging, it is on the way to develop the system which enables agents judge if its own solution is worth to be informed for other agents by this algorithm. Consequently a novel bio-inspired optimization algorithm BTO-PJ (Bee Total Optimization with Personal Judgment) is proposed. The BTO-PJ is a multi-agent system based on foraging activities of honey bees. In order to examine characteristics of the BTO-PJ, we apply it to the Traveling Salesperson Problem (TSP).
This paper presents the ACO algorithm with divide-and-conquer technique (We termed it DAS) for the TSP. Consequently, some numerical simulations demonstrate that the DAS is able to generate shorter tours than the conventional SA (Simulated Annealing), 2-opt and AS techniques.
The self-organizing mixture models (SOMMs) were proposed as an expectation-maximization (EM) algorithm that yields topology preserving maps of data based on probabilistic mixture models.Compared to self-organizing maps, the SOMM algorithm has a clear interpretation: it maximizes the sum of data log likelihood and a penalty term that enforces self-organization. The object of this paper is to extend the SOMM algorithm to deal with multivariate time series. The standard SOMM algorithm assumes that the data are independent and identically distributed samples. However, the i.i.d. assumption is clearly inappropriate for time series. In this paper we propose the extension of the SOMM algorithm for multivariate time series, which we call self-organizing hidden Markov models (SOHMMs), by assuming that the time series is generated by hidden Markov models (HMMs).
Two different types of transshipments which are called preventive or emergency transshipment, have been studied separately in the inventory distribution problem. The transshipment in inventory distribution system is important to improve customer service and reduce total cost. In this paper the inventory distribution problem using both transshipments is formulated as the stochastic programming problem in which customer demand is defined as random variables. The algorithm using L-shaped method is developed and the numerical experiments show that the proposed algorithm is quite efficient. Finally, the advantage using both of transshipments is shown.
Many Japanese people are hard to speak English smoothly because English words do not appear in their brain immediately. This is because Japanese students have been learned English by memorizing each word with Japanese. Therefore, they need to translate a word from Japanese into English whenever they speak English. In this paper, an interface for acquiring English skills is proposed. Users of the interface can learn English sense by seeing their own and others' input histories to answer English grammatical questions. According to the experimental results, users of the proposed system could answer quickly with higher accuracy rates.
As we make a refreshing shampoo formulation, in order to achieve the required quality, we have to examine the combination of the algefacient density, washing ingredients and conditioning ingredients. The evaluation of the performance for the formulated refreshing shampoo substantially depends on the result of quantitative research, such as the survey to consumers and the formulator's experimental rule. Those methods, however, are inaccurate because it is based on subjective data. To resolve the problem, with applying Emotion Fractal Analysis Method (EFAM) to electroencephalogram signals, we analyzed emotions such as sensation of coolness, refreshment, comfort and relaxation. The results showed that the correlation coefficient between evaluation value of EFAM and subjective assessment of refreshment was very high: 0.99. Those results point to a possibility that EFAM may be objectively measure particular human emotions.
In this study, the acquisition of rule-type knowledge from field inspection data on highway bridges isenhanced by introducing an improvement to a traditional data mining technique. The new rough set theory approach helps in cases of exceptional and contradictory data, which in the traditional decision table reduction method are simply removed from analyses. There are numerous inconsistent data in real data owned and managed by a highway corporation in Japan. A new method is therefore proposed to solve the problem of data loss. The new method reveals some generally unrecognized decision rules in addition to generally accepted knowledge. Finally, a computer programs is developed to perform calculation routines, and some field inspection data on highway bridges is used to show the applicability of the proposed method.
Handwriting movement includes a lot of intelligence such as identity, personality, tacit or explicit skills and so on. Especially, identity authentication by biometric technologies is currently gaining popularity over traditional password based security systems. Online handwritten signature verification is a long time candidate in this area of research. In the present paper handwritten signature is considered as a behavioral biometric attribute produced by the dynamics of human hand movement. We propose an approach for improving verification accuracy from the analysis of reconstruction of the dynamics from multi-dimensional time series generated from online handwritten signature.The approach contains the proposal of a new similarity measure cross translation error (CTE) to measure similarity of local dynamics between two time series and an integration of the proposed measure with conventional Dynamic Time Warping (DTW) which belongs to global dynamics measure. The simulation results with a small scale generated data set and the benchmark data set used in Signature Verification Contest (SVC) 2004 show that the proposed measure is effective in detecting individuality from handwritten time series compared to the popular DTW based measures. The integrated approach is also promising for increasing verification accuracy.
The idea of therapeutic category is valid to ensure the safe use of drugs. For example, when physicians prescribe using prescription ordering system, matching the therapeutic category of prescribed drugs to patients' disease information is expected to prevent incorrect administrations. However, current therapeutic category is inadequate for this purpose, since it is based on not only drug efficacies but also other viewpoints such as active ingredients of drugs. Furthermore, therapeutic category numbers are determined based on Japan standard commodity classification (JSOC) numbers compiled in 1990. Because of its oldness, some drugs are not applicable to any category and are classified as “others”. In this study, we propose the method to classify drugs purely based on effect-efficacy information described in medical package inserts.