Network structure is effective model as dealing with the world represented by nodes and edges. Previous studies mainly focused on small networks, namely human network. However, large structural network data, such as Internet is emerging all over the world. And it has many interesting features including scale-free structure, small-world, and so on. In this paper, we propose the fuzzy clustering model for scale-free network represented as adjacency matrix. Our model’s novelty is based on negative degree correlation, membership matrix and distance to deal with scale-free structure network. We evaluate our proposed method against two major fuzzy clustering method on four real world data sets and show that our method outperforms them for three scale-free structure network data.
This paper investigates the performance of distribution clustering techniques. We use L<jats:sup>2</jats:sup> distance as a distance metric between two distributions. This distance is obtained by estimating the difference between the estimators of the two distributions. As the estimation methods of the density difference, we consider two methods. One is two-step method where first each of the two density functions is estimated by kernel method, and then calculate the differrence between them. The other is direct method where the least-squares density-difference estimation method is used. The clustering performance with each method is investigated through a series of computational experiments.
This paper focuses on the autonomous flight of a small glider robot. The authors present a steering control system with a fuzzy inference unit. The glider robot is required to passively increase the altitude by utilizing updrafts since it cannot generate the propulsion force. For this purpose, both the antecedent and consequent parts taking the human-operated steering control into account are introduced into the fuzzy inference unit. Through simulation experiments, we show that the glider robot is enabled to reach a distant destination.