Curriculum Vitaes
  1. Researcher Search Results
  2. TANAKA MAMORU

TANAKA MAMORU

  (田中 衞)

Profile Information

Affiliation
(名誉教授), 理工学部 情報理工学科, 上智大学
Degree
Doctor of Engineering(Keio University)
Researcher number
00146804
J-GLOBAL ID
200901062366095741
researchmap Member ID
5000064364
External link

Tanaka Mamoru (Non-member) was born on 20 September 1948 in Kanagawa, Japan. He received the B.E, M.E amd Ph.D degrees in Electrical Engineering from Keio Univesity, Yokohama, Japan in the year 1972,1974 and 1981 recpectively. In 1974 he joined the Circuit Development Department, Computer Engineering Department of Nippon Electric Company (NEC) where he was engaged in design and development of NEC's LSI computers. He resigned from the company and entered in the graduate school of Keio Univesity in 1978. In April 1981, he became an Associate Professor with the Department of Electrical Electronics Engineering of Sophia Univesity, Tokyo, Japan. He is now a Professor of Sophia University. He has done research in analysis of a large scale of networks and architectures of new LSI computers, Data Mining, Machine Learning, Neural Networks and Circuit Analysis. He is now interested in the synthesis of Retina Chips by Cellular Neural Networks. He was an Associate Editor of the IEEE Transactions on Circuit and System. He is a fellow member of IEICE.

Research Interests

 1

Research Areas

 1

Research History

 1

Papers

 129
More

Misc.

 76
  • Nakaguchi Toshiya, Jin'no Kenya, Tanaka Mamoru
    IEICE technical report. Nonlinear problems, 97(372) 71-78, Nov 8, 1997  
    This paper describes analysis of synchronization phenomena from a simple hysteresis network which consists three cells. Since the simple hysteresis network is a piecewise linear system, we can calculate the trajectory using exact solutions. We focus on the relation between time constants and bifurcations, and consider the condition of the system exhibiting various phenomena.
  • YOSHIGAE Takahisa, JIN'NO Kenya, TANAKA Mamoru, KIMURA Tetsuya, HARADA Takaaki
    IEICE technical report. Nonlinear problems, 97(53) 55-62, May 23, 1997  
    The system that manipulator is track to desired trajectory by using multilayer-feedforward neural network has been reported in Ref [1] The connection coefficients of the system are tuned on-line, and the system outputs the control torque for manipulator to track desired trajectry In this report, that torque is generated by the system's connection cofficients, so we consider the dynamics of the torques that is controled by the feedforward NN's weights And we simulate that recurrent neural network learn the dynamics of the torque.
  • Miyata Junichi, Jinno Kenya, Tanaka Mamoru
    IEICE technical report. Nonlinear problems, 97(53) 47-54, May 23, 1997  
    In this paper, we propose descrete time cellular neural networks with asymmetric A templates. We show that it can be used to display input image on FLCD(Ferroelectric Liquid Crystal Display) using area intensity method for each Red, Green, Blue. FLCD panel needs asymmetric dynamics for its asymmetric pixel elements.
  • JINNO KEN'YA, TANAKA MAMORU
    電子情報通信学会技術研究報告, 97(53(NLP97 13-38)) 89-96, May 23, 1997  
    In this report, we consider hyperchaos in a simple hystresis network( ab, SHN ) which includes only three control parameters : mutual connection, self-feedback and DC term. The cell of SHN has bistable state. monostable state and astable state. Therefore, the cell exhibits equilibria attractors and periodic attractors. but it does not exhibits non-periodic attractors. However, we have reported that SHN exhibits non-periodic attractors. In this report, we observe that SHN exhibits "hyperchaos". And we consider that the output sequence of "hyperchaos".
  • SHIMADA DAISUKE, JINNO KEN'YA, TANAKA MAMORU
    電子情報通信学会技術研究報告, 97(53(NLP97 13-38)) 39-46, May 23, 1997  
    For analysis of direct current and transition, we have to solve first order linear simultaneous equations on a large scale by using two methods that one is direct method represented by LU method, another is relaxation method represented by Gauss-Seidel method. Circuit-simulator(SDP) which is developed by our laboratory can apply either method. In this report, we introduce an analysis algorithm of SDP which can execute on X-Window system. Also, we propose water distribution pipe networks analysis by using SDP.
  • Togami Atsushi, Miyata Jun'ichi, Jin'no Kennya, Tanaka Mamoru, Shingu Toshiaki, Inoue Yuji
    IEICE technical report. Neurocomputing, 96(584) 391-398, Mar 18, 1997  
    In this article, we propose an area intensity method with Discrete Time Cellular Neural Network(DT-CNN). In order to get high quality output image, we apply the DT-CNN to area intensity method. It causes "Shape Effect" which loses image quality, we have a discussion how to take place, and how to do.
  • MITA SHINSUKE, SAITO HIROYUKI, JINNO KEN'YA, TANAKA MAMORU
    電子情報通信学会技術研究報告, 96(584(NC96 155-208)) 267-274, Mar 18, 1997  
    In this article, we consider an image processing by using DT-CNN. In order to get high quality output image, we apply the DT-CNN whose output is multi levels. Also, we propose a novel intensity conversion method.
  • Senga Satoshi, Saito Hiroyuki, Jinno Kenya, Tanaka Mamoru
    Proceedings of the IEICE General Conference, 1997(Sogo Pt 1) 68-68, Mar 6, 1997  
    離散時間型セルラーニューラルネットワーク(DTCNN)を用いたブロックマッチングにより文字抽出をし、マッチ度の分布の中から得られる最大値を競合ネットワークの数理モデルと共に, そのダイナミクスを用いて最大検出するための応用を述べる.
  • Korehisa Makoto, Jin'no Kenya, Tanaka Mamoru
    IEICE technical report. Neurocomputing, 96(511) 79-86, Feb 6, 1997  
    This article proposes an implementation circuit of a Simple Hystersis Network whose output function is a piecewise linear hystersis. This circuit is an improvement version of a voltage mode circuit, the novel circuit drives current mode. We can apply the SHN to A/D converter, Winner-Take-All circuits and so on In this article, we propose an implementation circuit of a hysteresis quantizen, and we simulate it.
  • JIN'NO Kenya, TANAKA Mamoru
    IEICE technical report. Neurocomputing, 96(511) 25-30, Feb 6, 1997  
    We have proposed a si1nple hysteresis network (ab. SHN) wThose connection coefficents have uniform value in our previous works. In this report, we propose a generalized hysteresis network (ab. GSHN) whose connection coeeficients are given by the correlation matrix of a bipolar vector We can clarify that the attractors are controled by the self feedback parameter and the input. We forcus on the case where the all attrctors are stable equilibria. We classify the equilibra and clarify the number of attractors and their domain of attraction. Based on the result, we consider the network whose connection coefficients don't have restriction such as the GSHN. And, we introduce the hysteresis quanitzer which is an application.
  • SAITO HIROYUKI, JINNO KEN'YA, TANAKA MAMORU
    電子情報通信学会技術研究報告, 96(272(CAS96 26-40)) 65-72, Sep 27, 1996  
    We have studyed about Discrete-Time Cellular Neural Network(DT-CNN). However we have not given a indication of what DT-CNN has functions as same as Continuous Time Cellular Neural Network(CT-CNN). In this report, we show some templates for DT-CNN which have the same functions as CT-CNN. Moreover, we have good results about extracting characters and we show it. And we report that this method can also recognize characters.
  • JINNO KEN'YA, TANAKA MAMORU
    電子情報通信学会技術研究報告, 96(272(CAS96 26-40)) 81-88, Sep 27, 1996  
  • KOREHISA MAKOTO, JINNO KEN'YA, TANAKA MAMORU
    電子情報通信学会技術研究報告, 96(272(CAS96 26-40)) 73-79, Sep 27, 1996  
    We consider continuous-time analog cellular neural network incorporates hysteresis for VLSI. This cell can oscillate by itself, so this network generates various attractors. This network available for application of A/D converter, Winner Take All circuit and so on. In this paper, we implemented simple network circuit, and simulated cell oscillation by itself.
  • HATTORI TAIZO, JINNO KEN'YA, TANAKA MAMORU
    電子情報通信学会技術研究報告, 96(94(NLP96 34-45)) 41-48, Jun 14, 1996  
    The CNN (cellular neural network) that consists of locally connected neurons realizes the silicon retina. This paper describes CNN state equations to perform dynamic depth extraction for binocular stereo vision. The transmitter compress stereo images by constructure and area gradation quantization. Then the receiver reconstruct stereo images, and extracting parallax. The correspondence problem between left and right images is solved by block matching process of depth extraction system.
  • TANAKA MAMORU, JINNO KEN'YA, KANAYA MITSUHISA, TAKEDA KEN'ICHI
    電子情報通信学会技術研究報告, 96(73(NLP96 11-33)) 133-140, May 25, 1996  
    This paper describes two Sophia methods by which digital code theory developed as Hamming distance over Galois field GF(2) is expanded to analog code theory developed as Euclid distance in Real field R. The one is a method by which an incident matrix A corresponds to a parity check matrix H and the other is a method by which the parity check matrix H in GF(2) can be used directly as an analog parity check matrix. The coding and decoding methods are described by CNN state equations.
  • JINNO KEN'YA, TANAKA MAMORU
    電子情報通信学会技術研究報告, 96(73(NLP96 11-33)) 127-132, May 25, 1996  
    In our previous studies, we have analyzed the simple hysteresis network which has only two parameters. For the simple hysteresis network, we have clarified the number of attractors and their domain of attraction. In this article, we propose row uniform hysteresis networks and analyze their attractors. Also, we propose an area gradation system which is based on the row uniform hysteresis networks.
  • IKEGAMI Munemitsu, TANAKA Mamoru
    IEICE technical report. Nonlinear problems, 95(605) 55-62, Mar 26, 1996  
    This paper describes about the discrete time cellular neural network(DTCNN) and its application. A cell of the DTCNN has a staircase function as a output function and works at the discrete time. We can solve many optimizing problems by the DTCNN. As Its application, this paper describes the color quantization for the static images and moving images by the DTCNN for the pseudo color representation.
  • TAKEDA Kenichi, KANAYA Mitsuhisa, TANAKA Mamoru
    IEICE technical report. Nonlinear problems, 95(605) 39-46, Mar 26, 1996  
    This paper describes analog coding and decoding dynamics of orthogonal projection type of neural network(OPNN) with winner-take-all(WTA) operation. And, OPNN which has sparse connection on hidden layer contributed to implementation of hardware.
  • Miyata Junichi, Ikegami Munemitu, Tanaka Mamoru
    Proceedings of the IEICE General Conference, 1996(2) 5-5, Mar 11, 1996  
  • Ikegami Munemitsu, Tanaka Mamoru
    Proceedings of the IEICE General Conference, 1996(2) 101-101, Mar 11, 1996  
  • Tanaka Mamoru
    Proceedings of the IEICE General Conference, 1996 545-546, Mar 11, 1996  
  • Korehisa Makoto, Pham Cong-Kha, Tanaka Mamoru
    Proceedings of the IEICE General Conference, 1995 79-79, Mar 27, 1995  
  • KANAYA Mitsuhisa, TANAKA Mamoru
    Proceedings of the IEICE General Conference, 1995(1) 76-76, Mar 27, 1995  
  • Tanaka Mamoru
    IEICE technical report. Nonlinear problems, 94(259) 77-90, Sep 24, 1994  
    This paper describes CNN(cellular neural network)state equations to perform direct dynamic image coding-decoding and dynamic depth extraction for binocular stereo vision.The quantization for the funneling information is done by parallel neurons without the use of multiple bit analog-to-digital converters.And,the correspondence problem for extracting patallax between left and right image can be solved by pattern matching for analog images reconstructed from the transmitted binary syndrome.We make the simulations for moving image coding-decoding and for constructing the binocular stereo vision.
  • Kanaya Mitsuhisa, Tanaka Mamoru
    IEICE technical report. Nonlinear problems, 94(259) 69-76, Sep 24, 1994  
    In this paper we propose a novel method based on the local current comparison method,of which feature is to use cellular neural networks to find the maximum local current and to search the path,for planning the driving paths of multi-robot,and show some simulation results of it.The local current comparison method is deeply related to node analysis in neighbor,and this method suits to be constructed as the hardware on the analog-digital hybrid chip.Its basic principal is based on analog dynamics and it will make the plans so fast that it can be expected realization of the system which can generate plans for robots driving comparatively fast in real-time.
  • TANAKA Mamoru
    Journal of the Japan Society for Simulation Technology, 2(2) 87-97, Jul 15, 1983  

Books and Other Publications

 3

Presentations

 241
More

Professional Memberships

 2

Research Projects

 5
To the list screen