An emergent method for self-organizing a new small-world (SW) network with less average path-length than those of conventional small-world networks is proposed. The method is inspired by Ant- Colony Optimization (ACO), which is based on a pheromone trail formation by a collective behavior of ants. The resultant network architecture includes some "star" structure nodes with many degrees and other peripheral nodes with a few degrees. We called it a multi-star network, and analyzed its corresponding property of an n-star network theoretically and experimentally, comparing with typical conventional complex networks such as a random graph, WS (Watts-Stragatz) model and BA (Barabasi-Albert) model. We found that the new small-world network has an interesting property compared to other conventional complex networks, and it seems to reflect a real-world phenomenon such as the behavior of some "star" persons and their followers in a Twitter community and/or a Social Networking Service (SNS).

As the location-acquisition technologies become increasingly pervasive, tracking the movement of objects from trajectory datasets are more and more available. As a result, discovering frequent movement patterns from such a dataset has recently gained great interest. However, trajectory dataset is usually large in volume and exceeds the computation capacity of traditional centralized technologies. We propose a new approach to discovering patterns over a massive data set based on distributed storage and computing. We apply the proposed approach to different real-world datasets in different conditions. We also discuss the results and possible future research directions.

This paper proposes a new approach integrating the modeling of moving persons from sensor data and agent-based simulation for indoor layout design. First, we applied this framework to the prevention of children's accidents. Our model focuses on interaction between indoor objects and children to estimate the risk of indoor accidents. Next, it was used for workflow analysis and layout design in an operation room. Finally, We discuss the agent-based simulation of multiple persons moving in public spaces and its application to evaluating information presentation for guidance.

One practical inconvenience in frequent pattern mining is that it often yields a flood of common or uninformative patterns, and thus we should carefully adjust the minimum support. To alleviate this inconvenience, based on FP-growth, this paper proposes RP-growth, an efficient algorithm for top-k mining of discriminative patterns which are highly relevant to the class of interest. RP-growth conducts a branch-and-bound search using anti-monotonic upper bounds of the relevance scores such as F-score and 2, and the pruning in branch-and-bound search is successfully translated to minimum support raising, a standard, easy-to-implement pruning strategy for top-k mining. Furthermore, by introducing the notion of weakness and an additional, aggressive pruning strategy based on weakness, RP-growth efficiently find k patterns of wide variety and high relevance to the class of interest.

This paper presents a novel application of Graph Sequence Mining. Graph Sequence Mining is a method for finding common changes from sequences of graphs, and it have been applied to social networks, co-author networks, citation networks and so on. In this paper, we apply Graph Sequence Mining to a dependency parser based on transitions to mine rules to rewrite states in the parser.

We propose a novel efficient on-line algorithm for extracting frequent subsequences from a multiple-data stream. This algorithm solves the important problem that a large amount of memories are suddenly consumed when bursty arrivals occurs in a data stream. For an on-line algorithm, suppressing memory consumption is very important, thus, an on-line algorithm often takes a form of an approximation algorithm, where an error ratio is guaranteed to be lower than a user-specified threshold value. Our algorithm is based on an extended version [6] of LOSSY COUNTING Algorithm. The proposed on-line algorithm firstly limit the available memory to a given fixed space. Whenever it consumes all of the given memory space, it expires lowest frequency candidates of frequent sequences from the memory and stored instead new candidates which arrives in a data stream. We prove that the proposed algorithm have no false negatives under some conditions, and also have some other properties such as robustness.

The ambient networks need a function to select proper interactions in order to satisfy users. In this research, we propose a method to find the proper interactions to sleepy or sleeping users in a working space (e.g., lab and office) by reinforcement learning. This method controls a lamp, an aroma device and a music device and finds an interaction sequence that lets the users awake comfortably. We conducted an experiment to verify the validity of this method. In addition, we considered if measuring and analyzing a human vital sign (finger pulse) enable to get a user's condition.