論文 - 田中 剛平
-
Robustness and fragility in coupled oscillator networks under targeted attacks 査読あり 国際誌
Tianyu Yuan, Kazuyuki Aihara, Gouhei Tanaka
PHYSICAL REVIEW E 95 ( 1 ) 012315 - 012315 2017年01月
担当区分：最終著者, 責任著者 記述言語：英語 掲載種別：研究論文（学術雑誌） 出版者・発行元：AMER PHYSICAL SOC
The dynamical tolerance of coupled oscillator networks against local failures is studied. As the fraction of failed oscillator nodes gradually increases, the mean oscillation amplitude in the entire network decreases and then suddenly vanishes at a critical fraction as a phase transition. This critical fraction, widely used as a measure of the network robustness, was analytically derived for random failures but not for targeted attacks so far. Here we derive the general formula for the critical fraction, which can be applied to both random failures and targeted attacks. We consider the effects of targeting oscillator nodes based on their degrees. First we deal with coupled identical oscillators with homogeneous edge weights. Then our theory is applied to networks with heterogeneous edge weights and to those with nonidentical oscillators. The analytical results are validated by numerical experiments. Our results reveal the key factors governing the robustness and fragility of oscillator networks.
-
Complex-Valued Neural Networks for Wave-Based Realization of Reservoir Computing. 査読あり
Akira Hirose, Seiji Takeda, Toshiyuki Yamane, Daiju Nakano, Shigeru Nakagawa, Ryosho Nakane, Gouhei Tanaka
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) 10637 449 - 456 2017年
担当区分：最終著者 記述言語：英語 掲載種別：研究論文（国際会議プロシーディングス） 出版者・発行元：Springer Verlag
In this paper, we discuss the significance of complex-valued neural-network (CVNN) framework in energy-efficient neural networks, in particular in wave-based reservoir networks. Physical-wave reservoir networks are highly enhanced by CVNNs. From this viewpoint, we also compare the features of reservoir computing and other architectures.
DOI： 10.1007/978-3-319-70093-9_47
その他リンク： https://dblp.uni-trier.de/db/conf/iconip/iconip2017-4.html#HiroseTYNNNT17
-
Simulation Study of Physical Reservoir Computing by Nonlinear Deterministic Time Series Analysis 査読あり
Toshiyuki Yamane, Seiji Takeda, Daiju Nakano, Gouhei Tanaka, Ryosho Nakane, Akira Hirose, Shigeru Nakagawa
NEURAL INFORMATION PROCESSING, ICONIP 2017, PT I 10634 639 - 647 2017年
記述言語：英語 掲載種別：研究論文（国際会議プロシーディングス） 出版者・発行元：SPRINGER INTERNATIONAL PUBLISHING AG
We investigate dynamics of physical reservoir computing by numerical simulations. Our approach is based on nonlinear deterministic time series analysis such as Takens’ theorem and false nearest neighbor methods. We show that this approach is useful for efficient design and implementation of physical reservoir computing systems where only partial information of the reservoir state is accessible. We take nonlinear laser dynamics subject to time delay as physical reservoir and show that the size of physical reservoir can be estimated by these method.
-
Waveform Classification by Memristive Reservoir Computing 査読あり
Gouhei Tanaka, Ryosho Nakane, Toshiyuki Yamane, Seiji Takeda, Daiju Nakano, Shigeru Nakagawa, Akira Hirose
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) 10637 457 - 465 2017年
担当区分：筆頭著者, 最終著者 記述言語：英語 掲載種別：研究論文（国際会議プロシーディングス） 出版者・発行元：Springer Verlag
Reservoir computing is one of the computational frameworks based on recurrent neural networks for learning sequential data. We study the memristive reservoir computing where a network of memristors, instead of recurrent neural networks, provides a nonlinear mapping from input sequential signals to high-dimensional spatiotemporal dynamics. First we formulate the circuit equations of the memristive networks and describe the simulation methods. Then we use the memristive reservoir computing for solving a waveform classification problem. We demonstrate how the classification ability depends on the number of reservoir outputs and the variability of the memristive elements. Our methods are useful for finding a better architecture of the memristive reservoir under the inevitable element variability when implemented with nano/micro-scale devices.
-
Parameter Scaling for Epidemic Size in a Spatial Epidemic Model with Mobile Individuals 査読あり 国際誌
Chiyori T. Urabe, Gouhei Tanaka, Kazuyuki Aihara, Masayasu Mimura
PLOS ONE 11 ( 12 ) e0168127 2016年12月
記述言語：英語 掲載種別：研究論文（学術雑誌） 出版者・発行元：PUBLIC LIBRARY SCIENCE
In recent years, serious infectious diseases tend to transcend national borders and widely spread in a global scale. The incidence and prevalence of epidemics are highly influenced not only by pathogen-dependent disease characteristics such as the force of infection, the latent period, and the infectious period, but also by human mobility and contact patterns. However, the effect of heterogeneous mobility of individuals on epidemic outcomes is not fully understood. Here, we aim to elucidate how spatial mobility of individuals contributes to the final epidemic size in a spatial susceptible-exposed-infectious-recovered (SEIR) model with mobile individuals in a square lattice. After illustrating the interplay between the mobility parameters and the other parameters on the spatial epidemic spreading, we propose an index as a function of system parameters, which largely governs the final epidemic size. The main contribution of this study is to show that the proposed index is useful for estimating how parameter scaling affects the final epidemic size. To demonstrate the effectiveness of the proposed index, we show that there is a positive correlation between the proposed index computed with the real data of human airline travels and the actual number of positive incident cases of influenza B in the entire world, implying that the growing incidence of influenza B is attributed to increased human mobility.
-
Wave-Based Neuromorphic Computing Framework for Brain-Like Energy Efficiency and Integration 査読あり
Yasunao Katayama, Toshiyuki Yamane, Daiju Nakano, Ryosho Nakane, Gouhei Tanaka
IEEE TRANSACTIONS ON NANOTECHNOLOGY 15 ( 5 ) 762 - 769 2016年09月
担当区分：最終著者 記述言語：英語 掲載種別：研究論文（学術雑誌） 出版者・発行元：IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC
We present a framework of wave-based neuromorphic computing aiming at brain-like capabilities and efficiencies with nanoscale device integration. We take advantage of the unique nature of elastic nondissipative wave dynamics in both computations and IO communications in between as a means to natively implement and execute neuromorphic computing functions such as weighted sum in a spatiotemporal manner. Lower bound analysis based on a memory model and wave group velocity scaling is provided for conceptual evaluations.
-
Exploiting Heterogeneous Units for Reservoir Computing with Simple Architecture 査読あり
Gouhei Tanaka, Ryosho Nakane, Toshiyuki Yamane, Daiju Nakano, Seiji Takeda, Shigeru Nakagawa, Akira Hirose
NEURAL INFORMATION PROCESSING, ICONIP 2016, PT I 9947 187 - 194 2016年
担当区分：筆頭著者 記述言語：英語 掲載種別：研究論文（国際会議プロシーディングス） 出版者・発行元：SPRINGER INT PUBLISHING AG
Reservoir computing is a computational framework suited for sequential data processing, consisting of a reservoir part and a read-out part. Not only theoretical and numerical studies on reservoir computing but also its implementation with physical devices have attracted much attention. In most studies, the reservoir part is constructed with identical units. However, a variability of physical units is inevitable, particularly when implemented with nano/micro devices. Here we numerically examine the effect of variability of reservoir units on computational performance. We show that the heterogeneity in reservoir units can be beneficial in reducing the prediction error in the reservoir computing system with a simple cycle reservoir.
-
Dynamics of Reservoir Computing at the Edge of Stability 査読あり
Toshiyuki Yamane, Seiji Takeda, Daiju Nakano, Gouhei Tanaka, Ryosho Nakane, Shigeru Nakagawa, Akira Hirose
NEURAL INFORMATION PROCESSING, ICONIP 2016, PT I 9947 205 - 212 2016年
記述言語：英語 掲載種別：研究論文（国際会議プロシーディングス） 出版者・発行元：SPRINGER INT PUBLISHING AG
We investigate reservoir computing systems whose dynamics are at critical bifurcation points based on center manifold theorem. We take echo state networks as an example and show that the center manifold defines mapping of the input dynamics to higher dimensional space. We also show that the mapping by center manifolds can contribute to recognition of attractors of input dynamics. The implications for realization of reservoir computing as real physical systems are also discussed.
-
Computational Performance of Echo State Networks with Dynamic Synapses 査読あり
Ryota Mori, Gouhei Tanaka, Ryosho Nakane, Akira Hirose, Kazuyuki Aihara
NEURAL INFORMATION PROCESSING, ICONIP 2016, PT I 9947 264 - 271 2016年
記述言語：英語 掲載種別：研究論文（国際会議プロシーディングス） 出版者・発行元：SPRINGER INT PUBLISHING AG
The echo state network is a framework for temporal data processing, such as recognition, identification, classification and prediction. The echo state network generates spatiotemporal dynamics reflecting the history of an input sequence in the dynamical reservoir and constructs mapping from the input sequence to the output one in the readout. In the conventional dynamical reservoir consisting of sparsely connected neuron units, more neurons are required to create more time delay. In this study, we introduce the dynamic synapses into the dynamical reservoir for controlling the nonlinearity and the time constant. We apply the echo state network with dynamic synapses to several benchmark tasks. The results show that the dynamic synapses are effective for improving the performance in time series prediction tasks.
-
A Hybrid Pooling Method for Convolutional Neural Networks 査読あり
Zhiqiang Tong, Kazuyuki Aihara, Gouhei Tanaka
NEURAL INFORMATION PROCESSING, ICONIP 2016, PT II 9948 454 - 461 2016年
担当区分：最終著者, 責任著者 記述言語：英語 掲載種別：研究論文（国際会議プロシーディングス） 出版者・発行元：SPRINGER INTERNATIONAL PUBLISHING AG
The convolutional neural network (CNN) is an effective machine learning model which has been successfully used in the computer vision tasks such as image recognition and object detection. The pooling step is an important process in the CNN to decrease the dimensionality of the input image data and keep the transformation invariance for preventing the overfitting problem. There are two major pooling methods, i.e. the max pooling and the average pooling. Their performances depend on the data and the features to be extracted. In this study, we propose a hybrid system of the two pooling methods to improve the feature extraction performance. We randomly choose one of them for each pooling zone with a fixed probability. We show that the hybrid pooling method (HPM) enhances the generalization ability of the CNNs in numerical experiments with the handwritten digit images.
-
Oscillation dynamics underlie functional switching of NF-κB for B-cell activation. 査読あり 国際誌
Inoue K, Shinohara H, Behar M, Yumoto N, Tanaka G, Hoffmann A, Aihara K, Okada-Hatakeyama M
NPJ systems biology and applications 2 16024 - 16024 2016年
-
Photonic Reservoir Computing Based on Laser Dynamics with External Feedback 査読あり
Seiji Takeda, Daiju Nakano, Toshiyuki Yamane, Gouhei Tanaka, Ryosho Nakane, Akira Hirose, Shigeru Nakagawa
NEURAL INFORMATION PROCESSING, ICONIP 2016, PT I 9947 222 - 230 2016年
記述言語：英語 掲載種別：研究論文（国際会議プロシーディングス） 出版者・発行元：SPRINGER INT PUBLISHING AG
Reservoir computing is a novel paradigm of neural network, offering advantages in low learning cost and ease of implementation as hardware. In this paper we propose a concept of reservoir computing consisting of a semiconductor laser subject to external feedback by a mirror, where input signal is supplied as modulation pattern of mirror reflectivity. In that system, non-linear interaction between optical field and electrons are enhanced in complex manner under substantial external feedback, leading to achieve highly nonlinear projection of input electric signal to output optical field intensity. It is exhibited that the system can most efficiently classify waveforms of sequential input data when operating around laser oscillation's effective threshold.
-
Detecting early warning signals for blackouts in power grids from the viewpoint of nonlinear dynamics 査読あり
Motoki Nagata, Yoshito Hirata, Naoya Fujiwara, Gouhei Tanaka, Hideyuki Suzuki, Kazuyuki Aihara
International Symposium on Nonlinear Theory and its Applications (NOLTA2015) 22 - 25 2015年12月
記述言語：英語 掲載種別：研究論文（国際会議プロシーディングス）
-
Dynamics of an HBV Model with Drug Resistance Under Intermittent Antiviral Therapy 査読あり
Ben-Gong Zhang, Gouhei Tanaka, Kazuyuki Aihara, Masao Honda, Shuichi Kaneko, Luonan Chen
INTERNATIONAL JOURNAL OF BIFURCATION AND CHAOS 25 ( 7 ) 2015年06月
記述言語：英語 掲載種別：研究論文（学術雑誌） 出版者・発行元：WORLD SCIENTIFIC PUBL CO PTE LTD
This paper studies the dynamics of the hepatitis B virus (HBV) model and the therapy regimens of HBV disease. First, we propose a new mathematical model of HBV with drug resistance, and then analyze its qualitative and dynamical properties. Combining the clinical data and theoretical analysis, we demonstrate that our model is biologically plausible and also computationally viable. Second, we demonstrate that the intermittent antiviral therapy regimen is one of the possible strategies to treat this kind of complex disease. There are two main advantages of this regimen, i.e. it not only may delay the development of drug resistance, but also may reduce the duration of on-treatment time compared with the long-term continuous medication. Moreover, such an intermittent antiviral therapy can reduce the adverse side effects. Our theoretical model and computational results provide qualitative insight into the progression of HBV, and also a possible new therapy for HBV disease.
-
Intermittent Androgen Suppression: Estimating Parameters for Individual Patients Based on Initial PSA Data in Response to Androgen Deprivation Therapy. 査読あり 国際誌
Yoshito Hirata, Kai Morino, Koichiro Akakura, Celestia S Higano, Nicholas Bruchovsky, Teresa Gambol, Susan Hall, Gouhei Tanaka, Kazuyuki Aihara
PloS one 10 ( 6 ) e0130372 2015年
記述言語：英語 掲載種別：研究論文（学術雑誌） 出版者・発行元：PUBLIC LIBRARY SCIENCE
When a physician decides on a treatment and its schedule for a specific patient, information gained from prior patients and experience in the past is taken into account. A more objective way to make such treatment decisions based on actual data would be useful to the clinician. Although there are many mathematical models proposed for various diseases, so far there is no mathematical method that accomplishes optimization of the treatment schedule using the information gained from past patients or "rapid learning" technology. In an attempt to use this approach, we integrate the information gained from patients previously treated with intermittent androgen suppression (IAS) with that from a current patient by first fitting the time courses of clinical data observed from the previously treated patients, then constructing the prior information of the parameter values of the mathematical model, and finally, maximizing the posterior probability for the parameters of the current patient using the prior information. Although we used data from prostate cancer patients, the proposed method is general, and thus can be applied to other diseases once an appropriate mathematical model is established for that disease.
-
Public opinion formation with the spiral of silence on complex social networks 査読あり
Takeuchi Daiki, Tanaka Gouhei, Fujie Ryo, Suzuki Hideyuki
Nonlinear Theory and Its Applications, IEICE 6 ( 1 ) 15 - 25 2015年
記述言語：英語 出版者・発行元：一般社団法人 電子情報通信学会
Public opinion is formed through social interactions of individuals. A mechanism behind the formation of a highly dominant public opinion is a sociological theory called the spiral of silence. Here we study opinion dynamics resulting from the spiral of silence, using an agent-based model with complex interaction networks. We show that an extremely dominant public opinion arises in the presence of a moderate proportion of neutrals and its dominance level is enhanced by social interactions. Furthermore, we demonstrate that a correlation between characteristics and social interactions of the individuals has a large influence on the opinion formation dynamics.
DOI： 10.1587/nolta.6.15
-
Wave-Based Reservoir Computing by Synchronization of Coupled Oscillators 査読あり
Toshiyuki Yamane, Yasunao Katayama, Ryosho Nakane, Gouhei Tanaka, Daiju Nakano
NEURAL INFORMATION PROCESSING, PT III 9491 198 - 205 2015年
記述言語：英語 掲載種別：研究論文（国際会議プロシーディングス） 出版者・発行元：SPRINGER INT PUBLISHING AG
We propose wave-based computing based on coupled oscillators to avoid the inter-connection bottleneck in large scale and densely integrated cognitive systems. In addition, we introduce the concept of reservoir computing to coupled oscillator systems for non-conventional physical implementation and reduction of the training cost of large and dense cognitive systems. We show that functional approximation and regression can be efficiently performed by synchronization of coupled oscillators and subsequent simple readouts.
-
Robustness of oscillatory behavior in correlated networks. 査読あり 国際誌
Takeyuki Sasai, Kai Morino, Gouhei Tanaka, Juan A Almendral, Kazuyuki Aihara
PloS one 10 ( 4 ) e0123722 2015年
記述言語：英語 掲載種別：研究論文（学術雑誌） 出版者・発行元：PUBLIC LIBRARY SCIENCE
Understanding network robustness against failures of network units is useful for preventing large-scale breakdowns and damages in real-world networked systems. The tolerance of networked systems whose functions are maintained by collective dynamical behavior of the network units has recently been analyzed in the framework called dynamical robustness of complex networks. The effect of network structure on the dynamical robustness has been examined with various types of network topology, but the role of network assortativity, or degree-degree correlations, is still unclear. Here we study the dynamical robustness of correlated (assortative and disassortative) networks consisting of diffusively coupled oscillators. Numerical analyses for the correlated networks with Poisson and power-law degree distributions show that network assortativity enhances the dynamical robustness of the oscillator networks but the impact of network disassortativity depends on the detailed network connectivity. Furthermore, we theoretically analyze the dynamical robustness of correlated bimodal networks with two-peak degree distributions and show the positive impact of the network assortativity.
-
Regularity and Randomness in Modular Network Structures for Neural Associative Memories 査読あり
Gouhei Tanaka, Toshiyuki Yamane, Daiju Nakano, Ryosho Nakane, Yasunao Katayama
2015 INTERNATIONAL JOINT CONFERENCE ON NEURAL NETWORKS (IJCNN) 1 - 7 2015年
担当区分：筆頭著者 記述言語：英語 掲載種別：研究論文（国際会議プロシーディングス） 出版者・発行元：IEEE
This study explores efficient structures of artificial neural networks for associative memories. Motivated by the real brain structure and the demand of energy efficiency in hardware implementation, we consider neural networks with sparse modular structures. Numerical experiments are performed to clarify how the storage capacity of associative memory depends on regularity and randomness of the network structures. We first show that a fully regularized network, suited for design of hardware, has poor recall performance and a fully random network, undesired for hardware implementation, yields excellent recall performance. For seeking a network structure with good performance and high implementability, we consider four different modular networks constructed based on different combinations of regularity and randomness. From the results of associative memory tests for these networks, we find that the combination of random intramodule connections and regular intermodule connections works better than the other cases. Our results suggest that the parallel usage of regularity and randomness in network structures could be beneficial for developing energy-efficient neural networks.
-
Node-wise robustness against fluctuations of power consumption in power grids 査読あり
Motoki Nagata, Naoya Fujiwara, Gouhei Tanaka, Hideyuki Suzuki, Eiichi Kohda, Kazuyuki Aihara
EUROPEAN PHYSICAL JOURNAL-SPECIAL TOPICS 223 ( 12 ) 2549 - 2559 2014年10月
記述言語：英語 掲載種別：研究論文（学術雑誌） 出版者・発行元：SPRINGER HEIDELBERG
We propose a new concept of node-wise robustness of power grids under variation of effective power in one load node using a mathematical model that takes into account the change in voltage and reactive power of load nodes. We employ the topology of the power grid in eastern Japan. We define the robustness as the threshold value of the effective power, above which the steady state loses its stability. We show that the robustness is highly heterogeneous among the load nodes. We find that the shortest path length from generators is most highly correlated with the robustness of the load nodes. We numerically demonstrate that the supply of reactive power enhances the robustness.
