Misc - TANAKA Gouhei
-
25pTF-2 Aging transitions in multi-layer oscillator networks
Morino Kai, Tanaka Gouhei, Aihara Kazuyuki
Meeting Abstracts of the Physical Society of Japan 66 ( 0 ) 296 2011
Language：Japanese Publisher：The Physical Society of Japan
-
23pGU-4 Recovery of global oscillations in coupled oscillator networks
Morino Kai, Tanaka Gouhei, Aihara Kazuyuki
Meeting Abstracts of the Physical Society of Japan 66 ( 0 ) 286 2011
Language：Japanese Publisher：The Physical Society of Japan
-
Numerical Analysis on Coupled Systems of Period-1 and Period-2 Limit Cycle Oscillators
OKADA Yusuke, TANAKA Gouhei, KOHNO Takashi, AIHARA Kazuyuki
IEICE technical report 109 ( 269 ) 203 - 208 2009.11
Language：Japanese Publisher：The Institute of Electronics, Information and Communication Engineers
Coupled oscillators have been widely studied, for instance, as a mathematical model to investigate the mechanism of emergence of biological rhythms generated by an ensemble of interacting biological cells. In this study, we focus on network dynamics in a coupled system of inhomogeneous dynamical oscillators. In particular, we consider the case in which each oscillator is given as the same dynamical system while heterogeneity of oscillators lies in different parameter values. Recently, Daido and Nakanishi (2004) investigated network dynamics in a coupled system of active elements with a parameter value corresponding to a limit cycle and inactive elements with a parameter value corresponding to an equilibrium, and studied aging transition which occurs as the proportion of the inactive elements is varied. Motivated by this work, we numerically investigate a diffusively coupled system of period-1 and period-2 limit cycle oscillators. We show phase diagrams where the coupling strength and the proportion of the period-1 oscillators are varied.
-
Backpropagation Learning Algorithm for Multi layer Phasor Neural Networks
Gouhei Tanaka, Kazuyuki Aihara
NEURAL INFORMATION PROCESSING, PT 1, PROCEEDINGS 5863 484 - 493 2009
Language：English Publisher：SPRINGER-VERLAG BERLIN
We present a backpropagation learning algorithm for multi-layer feedforward phasor neural networks using a gradient descent method. The state of a phasor neuron takes a complex-valued state on the unit circle in the complex domain. Namely, the state can be identified only by its phase component because the amplitude component is fixed. Due to the circularity of the phase variable, phasor neural networks are useful to deal with periodic and multivalued variables. Under the assumption that the weight coefficients are complex numbers and the activation function is a continuous and differentiable function of a phase variable, we derive an iterative learning algorithm to minimize the output error. In each step of the algorithm, the weight coefficients are updated in the gradient descent direction of the error function landscape. The proposed algorithm is numerically tested in function approximation task. The numerical results suggest that the proposed method has a better generalization ability compared with the other backpropagation algorithm based on linear correction rule.
-
Estimation on effects of preventive measures reducing force of infection based on an epidemic model
TANAKA Gouhei, AIHARA Kazuyuki
SEISAN KENKYU 61 ( 6 ) 1081 - 1084 2009
Language：Japanese Publisher：Institute of Industrial Science The University of Tokyo
&nbsp;&nbsp;&nbsp;It is estimated that the Pandemic influenza A (H1N1) 2009 emerging all over the world in this year would further spread. Although production of pandemic vaccines is intensively promoted, its maximum amount is limited. Therefore, preventive measures other than the vaccination are also considered to be significant for decreasing the number of patients. In this study, under the assumption that the force of infection can be reduced by patients wearing a mask and staying at home, air ventilation, careful gargle and hand-wash etc., we estimate the effects of them using an epidemic model. For instance, in the case that the basic reproduction number is 1.4, reduction of the force of infection by 10% corresponds to approximately 16 million pandemic vaccines and results in decreasing the morbidity by about 25%. [This abstract is not included in the PDF]<br>
-
カオス工学は何を可能にするか
タナカ ゴウヘイ
78 ( 11 ) 1242 - 1245 2008.11
-
Complex-Valued Multistate Associative Memory with Nonlinear Multilevel Function
TANAKA Gouhei, AIHARA Kazuyuki
IEICE technical report 107 ( 478 ) 13 - 18 2008.02
Language：Japanese Publisher：The Institute of Electronics, Information and Communication Engineers
A complex-valued neural network can be used for multistate associative memory by quantizing a neuronal state into a multivalued state with an appropriate threshold function. The complex-signum function used in conventional multistate associative memory models is a transformation including a multilevel signum function in essence. In the present report, we propose a complex-valued threshold function based on nonlinear multilevel functions and show the recall performance of the multistate associative memory based on the proposed method. Numerical experiments clarify how the recall capability is influenced by the parameter controlling the nonlinearity of the multilevel function, the number of the stored patterns, and the number of quantized states. We also demonstrate gray-level image reconstruction with the proposed method.
-
MP-533 前立腺癌の間欠的ホルモン療法の数理モデルによる解析(一般演題ポスター,第94回日本泌尿器科学会総会)
島田 尚, 合原 一幸, 出田 亜位子, 津本 国親, 辻 繁樹, 平田 祥人, 鈴木 大慈, 田中 剛平, 武内 功, 北村 唯一
97 ( 2 ) 500 2006
Language：Japanese Publisher：一般社団法人 日本泌尿器科学会
-
Multistate associative memory with parametrically coupled map networks
Gouhei Tanaka, Kazuyuki Aihara
IEEE Asia-Pacific Conference on Circuits and Systems, Proceedings, APCCAS 2 1017 - 1020 2004.12
The present paper proposes two types of multistate associative memory models using circle maps coupled through a parameter in the individual map. Each network utilizes a circle map with a specific bifurcation property as its component and realizes self-organizing chaotic dynamics in a memory association. The performance of the proposed networks is compared with that of a conventional multistate neural network in multistate associative memory tests. © 2004 IEEE.
-
Analysis of Bursting Cell Model with 2D Maps
TANAKA Gouhei, AIHARA Kazuyuki
IEICE technical report. Circuits and systems 103 ( 334 ) 73 - 78 2003.09
Language：Japanese Publisher：The Institute of Electronics, Information and Communication Engineers
From the viewpoint of biological information processing, bursting responses of excitable cells may play an important role in the interaction of the cells. In several experiments and model simulations, two coupled bursting cells with electrical synapses often display in-phase and anti-phase burst synchronization depending on the coupling strength. To clarify an essential mechanism of the change between the two types of burst synchronization, this report investigates a coupled system of identical cells, each of which is described as a 2D bursting mapping. We show that the in-phase and anti-phase synchronized bursting solutions are stable for positive and negative coupling coefficients, respectively. The transition concerns with the change of local transverse stability of an invariant subspace. Moreover, the parameter region of bursting responses in a single cell is compared with that in the coupled cell system.
