論文 - 田中 剛平
-
Co-evolution dynamics of epidemic and information under dynamical multi-source information and behavioral responses
Xiao Hong, Yuexing Han, Gouhei Tanaka, Bing Wang
Knowledge-Based Systems 252 2022年09月
掲載種別：研究論文（学術雑誌）
In the absence of effective treatment programs and limited medical resources, multi-source information dynamically evolves with an epidemic and motivates people to adopt behavioral responses, which contributes much to reducing their infection risk and suppressing the epidemic spread. Here, we aim at studying the effects of dynamical multi-source information and behavioral responses on the co-evolution of epidemic and information in time-varying multiplex networks. We propose the UAU-SIS (Unaware–Aware–Unaware– Susceptible–Infected–Susceptible) model with time-varying self-awareness and behavioral responses. Under the framework of time-varying multiplex networks and with a Microscopic Markov Chain Approach (MMCA), we analytically derive the epidemic thresholds for the proposed model. Experimental results for artificial networks show that time-varying behavioral responses can effectively suppress the epidemic spread with an increased epidemic threshold, while time-varying self-awareness can only reduce the scale of epidemic spread. In addition, the role of dynamical multi-source information in suppressing epidemic spread is limited. When the information transmission rate is beyond a certain critical value or the information efficiency is low, it will no longer affect the epidemic spread. Detailed analysis on the co-evolution of epidemic and information has to consider the heterogeneity of individuals in obtaining multi-source information and taking behavioral responses. Only when many people can obtain multi-source information and take behavioral responses, time-varying self-awareness and behavioral responses have a great impact on suppressing epidemic spread. Furthermore, we apply our proposed framework to two typical real-world networks and find that the results on real-world networks are consistent with those on artificial networks. Thus, the proposed method is expected to provide helpful guidance for coping with the COVID-19 or future emerging epidemics.
-
Simulation platform for pattern recognition based on reservoir computing with memristor networks
Gouhei Tanaka, Ryosho Nakane
SCIENTIFIC REPORTS 12 ( 1 ) 2022年06月
担当区分：筆頭著者, 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：NATURE PORTFOLIO
Memristive systems and devices are potentially available for implementing reservoir computing (RC) systems applied to pattern recognition. However, the computational ability of memristive RC systems depends on intertwined factors such as system architectures and physical properties of memristive elements, which complicates identifying the key factor for system performance. Here we develop a simulation platform for RC with memristor device networks, which enables testing different system designs for performance improvement. Numerical simulations show that the memristor-network-based RC systems can yield high computational performance comparable to that of state-of-the-art methods in three time series classification tasks. We demonstrate that the excellent and robust computation under device-to-device variability can be achieved by appropriately setting network structures, nonlinearity of memristors, and pre/post-processing, which increases the potential for reliable computation with unreliable component devices. Our results contribute to an establishment of a design guide for memristive reservoirs toward the realization of energy-efficient machine learning hardware.
DOI： 10.1038/s41598-022-13687-z
その他リンク： https://dblp.uni-trier.de/db/journals/corr/corr2112.html#abs-2112-00248
-
Computational Efficiency of Multi-Step Learning Echo State Networks for Nonlinear Time Series Prediction. 査読あり
Takanori Akiyama, Gouhei Tanaka
IEEE Access 10 28535 - 28544 2022年
-
Multi-reservoir echo state networks with sequence resampling for nonlinear time-series prediction. 査読あり
Ziqiang Li, Gouhei Tanaka
Neurocomputing 467 115 - 129 2022年
-
2022 roadmap on neuromorphic computing and engineering. 査読あり
Dennis V. Christensen, Regina Dittmann, Bernabé Linares-Barranco, Abu Sebastian, Manuel Le Gallo, Andrea Redaelli, Stefan Slesazeck, Thomas Mikolajick, Sabina Spiga, Stephan Menzel, Ilia Valov, Gianluca Milano, Carlo Ricciardi, Shi-Jun Liang, Feng Miao, Mario Lanza, Tyler J. Quill, Scott T. Keene, Alberto Salleo, Julie Grollier, Danijela Markovic, Alice Mizrahi, Peng Yao, J. Joshua Yang, Giacomo Indiveri, John Paul Strachan, Suman Datta, Elisa Vianello, Alexandre Valentian, Johannes Feldmann, Xuan Li, Wolfram H. P. Pernice, Harish Bhaskaran, Steve B. Furber, Emre Neftci, Franz Scherr, Wolfgang Maass 0001, Srikanth Ramaswamy, Jonathan Tapson, Priyadarshini Panda, Youngeun Kim, Gouhei Tanaka, Simon Thorpe, Chiara Bartolozzi, Thomas A. Cleland, Christoph Posch, Shih-Chii Liu, Gabriella Panuccio, Mufti Mahmud, Arnab Neelim Mazumder, Morteza Hosseini, Tinoosh Mohsenin, Elisa Donati, Silvia Tolu, Roberto Galeazzi, Martin Ejsing Christensen, Sune Holm, Daniele Ielmini, N. Pryds
Neuromorphic Computing and Engineering 2 ( 2 ) 22501 - 22501 2022年
-
Spin waves propagating through a stripe magnetic domain structure and their applications to reservoir computing 査読あり
Ryosho Nakane, Akira Hirose, Gouhei Tanaka
Physical Review Research 3 ( 3 ) 2021年09月
担当区分：最終著者 掲載種別：研究論文（学術雑誌）
Spin waves propagating through a stripe domain structure and reservoir computing with their spin dynamics have been numerically studied, focusing on the relation between physical phenomena and computing capabilities. Our system utilizes a spin-wave-based device that has a continuous magnetic garnet film and one-input/72-output electrodes on top. To control spatially distributed spin dynamics, a stripe magnetic domain structure and amplitude-modulated triangular input waves were used. The spatially arranged electrodes detected spin vector outputs with various nonlinear characteristics that were leveraged for reservoir computing. By moderately suppressing nonlinear phenomena, our system achieves 100% prediction accuracy in temporal exclusive-OR problems with a delay step up to 5. At the same time, it shows perfect inference in delay tasks with a delay step more than 7 and its memory capacity has a maximum value of 21. This study demonstrated that our spin-wave-based reservoir computing has a high potential for edge-computing applications and also can offer a rich opportunity for further understanding the underlying nonlinear physics.
-
A Numerical Exploration of Signal Detector Arrangement in a Spin-Wave Reservoir Computing Device. 査読あり
Takehiro Ichimura, Ryosho Nakane, Gouhei Tanaka, Akira Hirose
IEEE Access 9 72637 - 72646 2021年
掲載種別：研究論文（学術雑誌）
This paper studies numerically how the signal detector arrangement influences the performance of reservoir computing using spin waves excited in a ferrimagnetic garnet film. This investigation is essentially important since the input information is not only conveyed but also transformed by the spin waves into high-dimensional information space when the waves propagate in the film in a spatially distributed manner. This spatiotemporal dynamics realizes a rich reservoir-computational functionality. First, we simulate spin waves in a rectangular garnet film with two input electrodes to obtain spatial distributions of the reservoir states in response to input signals, which are represented as spin vectors and used for a machine-learning waveform classification task. The detected reservoir states are combined through readout connection weights to generate a final output. We visualize the spatial distribution of the weights after training to discuss the number and positions of the output electrodes by arranging them at grid points, equiangularly circular points or at random. We evaluate the classification accuracy by changing the number of the output electrodes, and find that a high accuracy (>90%) is achieved with only several tens of output electrodes regardless of grid, circular or random arrangement. These results suggest that the spin waves possess sufficiently complex and rich dynamics for this type of tasks. Then we investigate in which area useful information is distributed more by arranging the electrodes locally on the chip. Finally, we show that this device has generalization ability for input wave-signal frequency in a certain frequency range. These results will lead to practical design of spin-wave reservoir devices for low-power intelligent computing in the near future.
-
A Multi-Reservoir Echo State Network with Multiple-Size Input Time Slices for Nonlinear Time-Series Prediction. 査読あり
Ziqiang Li, Gouhei Tanaka
ICONIP (2) 13109 LNCS 28 - 39 2021年
担当区分：最終著者, 責任著者 掲載種別：研究論文（国際会議プロシーディングス）
A novel multi-reservoir echo state network incorporating the scheme of extracting features from multiple-size input time slices is proposed in this paper. The proposed model, Multi-size Input Time Slices Echo State Network (MITSESN), uses multiple reservoirs, each of which extracts features from each of the multiple input time slices of different sizes. We compare the prediction performances of MITSESN with those of the standard echo state network and the grouped echo state network on three benchmark nonlinear time-series datasets to show the effectiveness of our proposed model. Moreover, we analyze the richness of reservoir dynamics of all the tested models and find that our proposed model can generate temporal features with less linear redundancies under the same parameter settings, which provides an explanation about why our proposed model can outperform the other models to be compared on the nonlinear time-series prediction tasks.
DOI： 10.1007/978-3-030-92270-2_3
その他リンク： https://dblp.uni-trier.de/db/conf/iconip/iconip2021-2.html#LiT21
-
2021 Roadmap on Neuromorphic Computing and Engineering.
Dennis V. Christensen, Regina Dittmann, Bernabé Linares-Barranco, Abu Sebastian, Manuel Le Gallo, Andrea Redaelli, Stefan Slesazeck, Thomas Mikolajick, Sabina Spiga, Stephan Menzel, Ilia Valov, Gianluca Milano, Carlo Ricciardi, Shi-Jun Liang, Feng Miao, Mario Lanza, Tyler J. Quill, Scott T. Keene, Alberto Salleo, Julie Grollier, Danijela Markovic, Alice Mizrahi, Peng Yao, J. Joshua Yang, Giacomo Indiveri, John Paul Strachan, Suman Datta, Elisa Vianello, Alexandre Valentian, Johannes Feldmann, Xuan Li, Wolfram H. P. Pernice, Harish Bhaskaran, Emre Neftci, Srikanth Ramaswamy, Jonathan Tapson, Franz Scherr, Wolfgang Maass 0001, Priyadarshini Panda, Youngeun Kim, Gouhei Tanaka, Simon Thorpe, Chiara Bartolozzi, Thomas A. Cleland, Christoph Posch, Shih-Chii Liu, Arnab Neelim Mazumder, Morteza Hosseini, Tinoosh Mohsenin, Elisa Donati, Silvia Tolu, Roberto Galeazzi, Martin Ejsing Christensen, Sune Holm, Daniele Ielmini, N. Pryds
CoRR abs/2105.05956 2021年
掲載種別：研究論文（学術雑誌）
その他リンク： https://dblp.uni-trier.de/db/journals/corr/corr2105.html#abs-2105-05956
-
partial-FORCE: A fast and robust online training method for recurrent neural networks. 査読あり
Hiroto Tamura, Gouhei Tanaka
International Joint Conference on Neural Networks(IJCNN) 2021-July 1 - 8 2021年
担当区分：最終著者 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
Recurrent neural networks (RNNs) are helpful tools for modeling dynamical systems by neuronal populations, but efficiently training RNNs has been a challenging topic. In recent years, a recursive least squares (RLS) based method for modifying all the recurrent connections, called the full-Force method, has been gaining attention as a fast and robust online training rule. This method introduces a second network (called the teacher reservoir) during training to provide suitable target dynamics to all the hidden units of the task-performing network (called the student network). Thanks to the RLS-based approach, the full-FORCE method can be applied to training continuous-time networks and spiking neural networks. In this study, we propose a generalized version of the full-FORCE method: the partial-FORCE method. In the proposed method, only part of the student network neurons (called supervised neurons) is supervised by only part of the teacher reservoir neurons (called supervising neurons). As a result of this relaxation, the size of the student network and that of the teacher reservoir can be different, which is biologically plausible as a possible model of the memory transfer in the brain. Furthermore, we numerically show that the partial-FORCE method converges faster and is more robust against variations in parameter values and initial conditions than the full-FORCE method, even without the price of computational cost.
DOI： 10.1109/IJCNN52387.2021.9533964
その他リンク： https://dblp.uni-trier.de/db/conf/ijcnn/ijcnn2021.html#TamuraT21
-
Transfer-RLS method and transfer-FORCE learning for simple and fast training of reservoir computing models. 査読あり
Hiroto Tamura, Gouhei Tanaka
Neural Networks 143 550 - 563 2021年
-
Reservoir Computing with Diverse Timescales for Prediction of Multiscale Dynamics.
Gouhei Tanaka, Tadayoshi Matsumori, Hiroaki Yoshida, Kazuyuki Aihara
CoRR abs/2108.09446 2021年
担当区分：筆頭著者 掲載種別：研究論文（学術雑誌）
その他リンク： https://dblp.uni-trier.de/db/journals/corr/corr2108.html#abs-2108-09446
-
Processing-Response Dependence on the On-Chip Readout Positions in Spin-Wave Reservoir Computing. 査読あり
Takehiro Ichimura, Ryosho Nakane, Akira Hirose
Neural Information Processing - 28th International Conference 296 - 307 2021年
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
DOI： 10.1007/978-3-030-92238-2_25
その他リンク： https://dblp.uni-trier.de/db/conf/iconip/iconip2021-3.html#IchimuraNH21
-
Network structure-based interventions on spatial spread of epidemics in metapopulation networks 査読あり
Bing Wang, Min Gou, YiKe Guo, Gouhei Tanaka, Yuexing Han
PHYSICAL REVIEW E 102 ( 6 ) 2020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：AMER PHYSICAL SOC
Mathematical modeling of epidemics is fundamental to understand the mechanism of the disease outbreak and provides helpful indications for effectiveness of interventions for policy makers. The metapopulation network model has been used in the analysis of epidemic dynamics by taking individual migration between patches into account. However, so far, most of the previous studies unrealistically assume that transmission rates within patches are the same, neglecting the nonuniformity of intervention measures in hindering epidemics. Here, based on the assumption that interventions deployed in a patch depend on its population size or economic level, which have shown a positive correlation with the patch's degree in networks, we propose a metapopulation network model to explore a network structure-based intervention strategy, aiming at understanding the interplay between intervention strategy and other factors including mobility patterns, initial population, as well as the network structure. Our results demonstrate that interventions to patches with different intensity are able to suppress the epidemic spreading in terms of both the epidemic threshold and the final epidemic size. Specifically, the intervention strategy targeting the patches with high degree is able to efficiently suppress epidemics. In addition, a detrimental effect is also observed depending on the interplay between the intervention measures and the initial population distribution. Our study opens a path for understanding epidemic dynamics and provides helpful insights into the implementation of countermeasures for the control of epidemics in reality.
-
Comparing catch-up vaccination programs based on analysis of 2012-13 rubella outbreak in Kawasaki City, Japan 査読あり 国際誌
Chiyori T. Urabe, Gouhei Tanaka, Takahiro Oshima, Aya Maruyama, Takako Misaki, Nobuhiko Okabe, Kazuyuki Aihara
PLOS ONE 15 ( 8 ) e0237312 2020年08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
During the 2012-13 rubella outbreak in Japan, local governments implemented subsidy programs for catch-up vaccination to mitigate the rubella outbreak and prevent congenital rubella syndrome (CRS). In most local governments, to prevent CRS, eligible persons of the subsidy program were women who were planning to have a child and men who were partners of pregnant women. On the other hand, in Kawasaki City, unimmunized men aged 23-39 years were additionally included in the eligible persons, because they were included in an unimmunized men group resulting from the historical transition of the national routine vaccination in Japan. The number of rubella cases in the city decreased earlier than that in the whole Japan. First, in order to estimate the effect of the catch-up vaccination campaign in Kawasaki City on the epidemic outcome, we performed numerical simulations with a Susceptible-Vaccinated-Exposed-Infectious-Recovered (SVEIR) model incorporating real data. The result indicated that the catch-up vaccination campaign showed a beneficial impact on the early decay of the rubella cases. Second, we numerically compared several different implementation strategies of catch-up vaccinations under a fixed amount of total vaccinations. As a result, we found that early and intensive vaccinations are vital for significant reduction in the number of rubella cases and CRS occurrences. Our study suggests that mathematical models with epidemiological and social data can contribute to identifying the most effective vaccination strategy.
-
HP-ESN: Echo State Networks Combined with Hodrick-Prescott Filter for Nonlinear Time-Series Prediction. 査読あり
Ziqiang Li, Gouhei Tanaka
2020 International Joint Conference on Neural Networks(IJCNN) 1 - 9 2020年
担当区分：最終著者 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
Nonlinear time-series prediction is one of the challenging tasks in machine learning. Recurrent neural networks and their variants have been successful in such a task owing to its ability of storing past inputs in their dynamical states. Echo state networks (ESNs) are a special type of recurrent neural networks, which are capable of high-speed learning. To develop this computational scheme, we propose an HP-ESN method which combines ESNs with a preprocessing based on the Hodrick- Prescott (HP) filter. This filter extracts different components from a single time-series data. The extracted components are processed by ESNs. We show that the proposed method yields better prediction performance compared with other state-of- the-art ESN-based methods in prediction tasks with real-world time-series data. We also demonstrate that the computational performance depends on the setting of the smoothing parameter and the number of decompositions by the HP filter.
DOI： 10.1109/IJCNN48605.2020.9206771
その他リンク： https://dblp.uni-trier.de/db/conf/ijcnn/ijcnn2020.html#LiT20
-
Estimation and prediction of ellipsoidal molecular shapes in organic crystals based on ellipsoid packing. 査読あり 国際誌
Daiki Ito, Raku Shirasawa, Yoichiro Iino, Shigetaka Tomiya, Gouhei Tanaka
PloS one 15 ( 9 ) e0239933 2020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（学術雑誌）
Crystal structure prediction has been one of the fundamental and challenging problems in materials science. It is computationally exhaustive to identify molecular conformations and arrangements in organic molecular crystals due to complexity in intra- and inter-molecular interactions. From a geometrical viewpoint, specific types of organic crystal structures can be characterized by ellipsoid packing. In particular, we focus on aromatic systems which are important for organic semiconductor materials. In this study, we aim to estimate the ellipsoidal molecular shapes of such crystals and predict them from single molecular descriptors. First, we identify the molecular crystals with molecular centroid arrangements that correspond to affine transformations of four basic cubic lattices, through topological analysis of the dataset of crystalline polycyclic aromatic molecules. The novelty of our method is that the topological data analysis is applied to arrangements of molecular centroids intead of those of atoms. For each of the identified crystals, we estimate the intracrystalline molecular shape based on the ellipsoid packing assumption. Then, we show that the ellipsoidal shape can be predicted from single molecular descriptors using a machine learning method. The results suggest that topological characterization of molecular arrangements is useful for structure prediction of organic semiconductor materials.
-
Deep Echo State Networks with Multi-Span Features for Nonlinear Time Series Prediction. 査読あり
Ziqiang Li, Gouhei Tanaka
2020 International Joint Conference on Neural Networks(IJCNN) 1 - 9 2020年
担当区分：最終著者 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
Nonlinear time-series prediction is one of the challenging topics in machine learning due to complex non-stationarity in the temporal dynamics. Many recurrent neural network models have been proposed for enhancing the prediction accuracy in time-series prediction tasks. Echo state networks (ESNs) are a variant of recurrent neural networks, which have great potential for addressing machine learning tasks with a very low learning cost. However, the existing ESN-based models have used only single-span features to our best knowledge. In this study, we propose two deep ESN models incorporating multi-span features to improve the prediction performance. We show that the two deep ESN models yield better prediction performance compared to the other state-of-the-art ESN-based methods in benchmark time-series prediction tasks with three models: the Lorenz system, the Mackey-Glass system, and the NARMA-10 system. Our analyses illustrate that deeper structures decrease the multicollinearity of the extracted features and thus contribute to improved performance. The presented results suggest that the proposed models contribute to the development of artificial intelligence for temporal information processing.
DOI： 10.1109/IJCNN48605.2020.9207401
その他リンク： https://dblp.uni-trier.de/db/conf/ijcnn/ijcnn2020.html#LiT20a
-
Time delay reservoir computing with VCSEL
Jean Benoit Héroux, Gouhei Tanaka, Toshiyuki Yamane, Naoki Kanazawa, Ryosho Nakane, Hidetoshi Numata, Seiji Takeda, Akira Hirose, Daiju Nakano
Proceedings of SPIE - The International Society for Optical Engineering 11299 2020年
掲載種別：研究論文（国際会議プロシーディングス）
© COPYRIGHT SPIE. Downloading of the abstract is permitted for personal use only. Neural networks in which the interconnections between the nodes are randomly aßigned are promising for the realization of neuromorphic devices in which the resource requirements for training are lower than for a fully deterministic system. Reservoir computing is a claß of recurrent network for which the input and internal weights are random and fixed over time, and only the output weights are trained via a linear regreßion. In this work, we review the recent work on photonic reservoirs and describe our recent results on the implementation of a single node system based on multi-mode optical interconnect technology developed for high channel density and low power data transfer applications. We discuß the potential advantages of this approach for the realization of a photonic cluster of reservoirs.
DOI： 10.1117/12.2544981
-
Two-Step FORCE Learning Algorithm for Fast Convergence in Reservoir Computing. 査読あり
Hiroto Tamura, Gouhei Tanaka
Artificial Neural Networks and Machine Learning - ICANN 2020 - 29th International Conference on Artificial Neural Networks 12397 LNCS 459 - 469 2020年
担当区分：最終著者 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
Reservoir computing devices are promising as energy-efficient machine learning hardware for real-time information processing. However, some online algorithms for reservoir computing are not simple enough for hardware implementation. In this study, we focus on the first order reduced and controlled error (FORCE) algorithm for online learning with reservoir computing models. We propose a two-step FORCE algorithm by simplifying the operations in the FORCE algorithm, which can reduce necessary memories. We analytically and numerically show that the proposed algorithm can converge faster than the original FORCE algorithm.
DOI： 10.1007/978-3-030-61616-8_37
その他リンク： https://dblp.uni-trier.de/db/conf/icann/icann2020-2.html#TamuraT20
