Non-negative Matrix Factorization (NMF) is a basic method for decomposing matrices composed of only nonnegative values and has been utilized in various fields including air pollution analysis. However, based on the least square principle, NMF is easily influenced by noise. This paper proposes a robust NMF model by introducing noise rejection mechanism of noise fuzzy clustering with the goal of eliminating the influence of noise observation. Robust estimation is realized by estimating the degree of belongingness of each observation unit to noise clusters, which contributes to reducing the influences of noise in matrix decomposition. The characteristics of the proposed method are demonstrated in a toy example with an artificial data set followed by a task of air pollutant measurement analysis.

Collaborative data clustering is a promising approach for extracting intrinsic cluster structures from distributed databases keeping personal privacy, where the goal of collaborative analysis is to find richer information rather than independent analysis of each database. In this research, a novel privacy preserving linear fuzzy clustering model is proposed by enhancing the scheme of k-Means-type model into Fuzzy c-Lines (FCL) in conjunction with cryptographic calculation. The element-wise clustering criterion enables to derive local principal component vectors in each data sources without handling fuzzy scatter matrices. The characteristics of the proposed method are demonstrated in an experiment with an artificial data set followed by an application to analysis of human behavior from sensor data.

Time-series data is data that contains information about time-varying phenomena, and it has a wide range of applications. Clustering is one of the data analysis methods to analyze large complex time-series data and extract their features. The important issues in clustering time-series data is the selection of a suitable dissimilarity and the selection of a suitable clustering algorithm. In this paper, we propose new clustering methods to handle imbalanced time-series data by introducing the concept of size-control into the clustering methods for time-series data. The proposed methods are constructed by extending k-medoids using dynamic time warping (DTW) for dissimilarity, k-medoids and k-shape using shape-based distance (SBD) for dissimilarity, which are typical methods for time-series data. The performance of the proposed methods is verified by numerical experiments using 12 datasets available in the UCR Time Series Classification Archive. From the numerical experiments, we confirmed that k-medoids with size control using DTW obtains the best cluster partition among the proposed methods.

Children can imitate adults’ actions with ease. An imitator who observed demonstrator’s action can produce a bodily movement which is supposed to be similar with it. Any theory and hypothesis of imitation is required to describe how the imitator recognizes the similarity or identity of actions, i.e., in what sense from the imitator’s point-of-view the produced and the observed action can be similar or identified. In this paper, we review four existing hypotheses on the mechanism of imitation. One of our criteria for the hypotheses in this review is the possibility of implementation and validation of the hypotheses in computer simulations or physical robots. A straightforward implementation without any additional ad-hoc assumption would be difficult, if the hypothesis to be implemented is flawed. Thus, the computer-based and/or physical implementation tests if a hypothesis accounts for the imitation mechanism sufficiently. This motivates us to critically review and clarify the existing literature on imitation in the sense of technical and theoretical plausibility of the existing hypotheses. We review the four existing hypotheses on imitation with some of them had been implemented in a form of computer simulations and/or robots. By pointing out the additional assumptions for the implementation, this review will reveal the latent requirements for an account for the imitation, that has not been addressed well. Lastly, we briefly propose our own account for imitation, in which bodily movements are characterized and identified on the basis of dynamical invariants under smooth transformations.

In dialogue systems, conveying the understanding results of user utterances is important because it enables the users to feel understood by the system. However, it is not clear which types of understanding results should be conveyed to users; some utterances may be offensive, for example, and some may be too commonsensical. In this paper, we explore the effect of conveying the understanding results of user utterances in a chat-oriented dialogue system by an experiment with human subjects. We created system utterances conveying various understanding results and then investigated which types of results were favorable and unfavorable through manual evaluation. We found that understanding results referring to a user’s internal state were unfavorable, while those related to user’s positive attributes were likely to be favorable. In addition, understanding results related to objective facts about users or general facts unrelated to users were favorable. These findings can be used as a guideline for constructing a dialogue system that conveys appropriate understanding results.

In order to evaluate effects of usage of floor-plan images (FPIs) in a rent prediction, we construct rent-prediction models with and without FPIs and compare them in terms of prediction error. Principal component analysis, Bag of Features (BoF), and Fisher Vector (FV) are used as feature extractors, and linear regression and LightGBM are used as a regressor. A rent prediction experiment with the LIFULL HOME’S dataset is performed. For the linear regression models, the experimental results suggest that usage of FPIs improves the means of prediction errors in some categories (combination of a prefecture and a floor-plan standard) and reduces the 95% confidence intervals of prediction errors for all categories. In addition, BoF achieves the best prediction accuracy in the three feature extractors for FPIs.