Entity

Time filter

Source Type


Inoue H.,Kure National College of Technology
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) | Year: 2014

Recently, multiple classifier systems have been used for practical applications to improve classification accuracy. Self-generating neural networks (SGNN) are one of the most suitable base-classifiers for multiple classifier systems because of their simple settings and fast learning ability. However, the computation cost of the multiple classifier system based on SGNN increases in proportion to the numbers of SGNN. In this paper, we propose a novel pruning method for efficient classification and we call this model a self-organizing neural grove (SONG). Experiments have been conducted to compare the SONG with bagging and the SONG with boosting, and support vector machine (SVM). The results show that the SONG can improve its classification accuracy as well as reducing the computation cost. ©Springer International Publishing Switzerland 2014. Source


Sugitani T.,Hiroshima University | Kubota S.,Hiroshima University | Toya A.,Kure National College of Technology | Xiao X.,Tianjin University | Kikkawa T.,Hiroshima University
IEEE Antennas and Wireless Propagation Letters | Year: 2013

A compact 4 ×: 4 planar ultrawideband (UWB) antenna array with the total size of 44,×:, 52.4 mm2 was developed for radar-based breast cancer detection system. The center frequency and the bandwidth of the antenna were 6 and 12.5 GHz, respectively. The breast phantom materials were developed to fit the characteristics of the measured human breast tissues. A quasi-three- dimensional confocal imaging was performed using the breast phantoms. It was confirmed that the compact 4 ×: 4 antenna array could detect a 5×: 5×: 5-mm3 tumor phantom in an inhomogeneous structure with a glandular phantom and resolve the two separate tumor phantoms, which were located at the depth of 23 mm with the spacing of 10 mm. © 2002-2011 IEEE. Source


Inoue H.,Kure National College of Technology
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) | Year: 2013

Recently, multiple classifier systems have been used for practical applications to improve classification accuracy. Self-generating neural networks (SGNN) are one of the most suitable base-classifiers for multiple classifier systems because of their simple settings and fast learning ability. However, the computation cost of the multiple classifier system based on SGNN increases in proportion to the numbers of SGNN. In this paper, we propose a novel pruning method for efficient classification and we call this model a self-organizing neural grove (SONG). Experiments have been conducted to compare the SONG with bagging and the SONG with boosting, the multiple classifier system based on C4.5, and support vector machine (SVM). The results show that the SONG can improve its classification accuracy as well as reducing the computation cost. Additionally, we investigate SONG's incremental learning performance. © Springer-Verlag 2013. Source


Mori S.,Kure National College of Technology
Precision Engineering | Year: 2015

Bessel beams can be used for high aspect ratio laser drilling and for eliminating the need to precisely position materials along the propagation direction during laser drilling and cutting. However, Bessel beams have side lobes that can damage the materials subjected to these beams. This paper discusses optical suppression of side lobes. A method is proposed to suppress these side lobes, and the method is based on interference between two Bessel beams with different wave vectors. The effectiveness of this method is confirmed both theoretically and experimentally by realizing a superposed Bessel beam; using a He-Ne laser (λ = 633 nm) and an annular binary aperture placed in front of a convex lens, this beam has a 1/e2 radius of 44 μm. © 2014 Elsevier Inc. Source


Yoshikawa Y.,Kure National College of Technology
Proceedings of the Asia and South Pacific Design Automation Conference, ASP-DAC | Year: 2013

A binding method in high-level synthesis for path delay testability is proposed in this paper. For a given scheduled data flow graph, the proposed method synthesizes a path delay testable RTL datapath and its controller. Every path in the datapath is two pattern testable with the controller if the path is activated in the functional operation, i.e., the path is not false path. Our experimental results show that the proposed method can synthesize such RTL circuits with small area overhead compared with that augmented by some DFT techniques such as scan design. © 2013 IEEE. Source

Discover hidden collaborations