Reorganizing knowledge in neural networks: an explanatory mechanismfor neural networks in data classification problems

We propose an explanatory mechanism for multilayered neural networks (NN). In spite of the effective learning capability as a uniform function approximator, the multilayered NN suffers from unreadability, i.e., it is difficult for the user to interpret or understand the "knowledge" that the NN has by looking at the connection weights and thresholds obtained by backpropagation (BP). This unreadability comes from the distributed nature of the knowledge representation in the NN. In this paper, we propose a method that reorganizes the distributed knowledge in the NN to extract approximate classification rules. Our rule extraction method is based on the analysis of the function that the NN has learned, rather than on the direct interpretation of connection weights as correlation information. More specifically, our method divides the input space into "monotonic regions" where a monotonic region is a set of input patterns that belongs to the same class with the same sensitivity pattern. Approximate classification rules are generated by projecting these monotonic regions.     

The evaluation of speckle contrast with variable speckle generator

Abstract— While laser projection has many advantages, there is a problem with speckle patterns generated as a result of interference of the laser beam and results in bad effects to observers. In 2010, a variable speckle generator, which produces an angular shift of incident light to the screen and generates variable speckle patterns, was suggested. In this study, the performance of a variable speckle generator by using a volume phase holographic beam shaper and scanning mirror was investigated in detail by evaluating both the objective and subjective speckle contrast. The morphology of the speckle pattern was also investigated when the variable speckle generator was activated. With a scanning VPH beam shaper, the objective speckle was effectively reduced because each point of the VPH beam shaper generated different speckle patterns and coherency among each pattern disappeared by using the scan process. On the other hand, subjective speckle was also dramatically reduced by changing the incident angle on the screen, which resulted in generating variable subjective speckle. It was also shown that the speckle reduction rate by using a variable speckle generator did not depend on the coherent length of a laser by evaluating the normalized speckle contrast against the angular shift on the screen.     

Performance evaluation of immunity-based statistical en-route filtering in wireless sensor networks

Statistical en-route filtering (SEF) schemes can detect and eliminate false data injection attacks in wireless sensor networks. However, SEF does not address the identification of the compromised nodes which are injecting false reports. Therefore, we have proposed an immunity-based SEF to identify compromised nodes and achieve earlier detection of false reports. In the proposed scheme, each node has a list of neighborhood nodes and assigns credibility to each neighboring node. Each node can update the credibility of a neighboring node based on the success or failure of filtering and communication, and can then use the updated credibility as the probability of the next communication. In this article, some simulation results show that the immunity-based SEF outperforms the original SEF.     

Cognitive,social and teaching presence in a virtual world and a text chat

Using a framework of cognitive, social, and teaching presence, the nature of learning experiences in a three-dimensional virtual world environment (Second Life) and a text-chat learning environment without visuals (TeachNet) were investigated. A mixed method of code frequencies, coherence graphs, interviews, and a survey was used. The results revealed that the TeachNet debates included more cognitive presence codes that indicate higher levels of cognitive processing than in SL debates. The teams were significantly different from each other in the collaboration style for developing arguments and in the ways to use utterances associated with cognitive, social, and teaching presences, and the groups’ collaboration style became more established with more experience with the tasks. The three critical factors–tool, tasks and group cohesion–that affect cognitive, teaching, and social presence are discussed.     

Stick-slip motion control based on cutter location data for an orthogonal-type robot

In this article, a new desktop orthogonal-type robot, which has the capacity of stick-slip motion control based on cutter location data, is presented for lapping small metallic molds with a curved surface. The robot consists of three single-axis devices with a high position resolution of 1 μm. A thin wooden stick tool with a ball-end shape is attached to the tip of the z-axis. In order to improve the lapping performance, a novel stick-slip motion control method is developed in the control system. The small stick-slip motion is orthogonally generated in the direction of the tool’s movement. The effectiveness of stick-slip motion control is examined through an actual lapping test of an LED lens cavity.     

New dissimilarity measure for recognizing noisy subsequence trees

Tree is a data structure used to express various objects such as semistructured data and genes. When objects are represented as trees, computing tree similarity is essential for pattern recognition and retrieval. This paper considers the noisy subsequence tree recognition problem whose purpose is to recognize the original tree, given its noisy subsequence tree. Previous research on this problem relied on constrained tree edit distance to measure the dissimilarity. However, the number of relabelings must be predetermined to compute it. This paper proposes a new dissimilarity measure for this problem. Our dissimilarity measure is obtained by counting the node edit operations included in the unit‐cost tree edit distance that contribute to the matching of node labels. The number of relabelings need not be specified to compute our dissimilarity measure. Moreover, our measure achieves more accurate recognition performance and faster execution speed than the constrained tree edit distance. Our measure is also useful to solve the tree inclusion problem which is the problem of deciding whether a tree includes another tree and shows the extent of approximate tree inclusion when a tree incompletely includes another tree. © 2011 Wiley Periodicals, Inc.     

Scale free interval graphs

Scale free graphs have attracted attention by their non-uniform structure that can be used as a model for various social and physical networks. In this paper, we propose a natural and simple random model for generating scale free interval graphs. The model generates a set of intervals randomly under a certain distribution, which defines a random interval graph. The main advantage of the model is its simpleness. The structure/properties of generated graphs are analyzable by relatively simple probabilistic and/or combinatorial arguments, which is different from many other models. Based on such arguments, we show for our random interval graph that its degree distribution follows a power law, and that it has a large average clustering coefficient.     

iOS application for quadrotor remote control

With the progress of electronics technology, the development of civilian UAV (unmanned aerial vehicle) applications becomes possible. In addition, smartphones have rapidly gained popularity and become very important due to the simple operability and mobility. Hence, there is a need to have an easy and flexible way to control a UAV using such technology. In this study, a remote controller using an iOS device is developed for a quadrotor to enable remote control with easy operations. Four basic programs for obtaining compass information, controlling a gimbal, autopilot function for return, and video preview function are developed and implemented for an iOS device. The basic functionalities of the programs are evaluated and confirmed through experiments using a quadrotor and an iOS device.     

Control of three-link underactuated manipulators using a switching method of fuzzy energy regions

In general, manipulators used for industry and in academic laboratories have actuators to drive each joint. On the other hand, underactuated manipulators handled by our research have some passive or free joints without actuators and brakes. We recently developed a switching method of fuzzy energy regions to control such manipulators. In such a method, it is necessary to design parameters related to energy regions and the gains of some partly stable controllers based on the computed torque method. Here, the switching method is applied for a three-link underactuated manipulator. We optimize such design parameters related to fuzzy energy regions by a genetic algorithm. The effectiveness of the present method is illustrated with some simulations. This work was presented in part at the 12th International Symposium on Artificial Life and Robotics, Oita, Japan, January 25–27, 2007