DSpace Collection:
http://hdl.handle.net/2115/20053
2024-03-29T07:56:11ZA Simple and Versatile Evaluation Method of Thermal Stability of NI HTS Magnets
http://hdl.handle.net/2115/91228
Title: A Simple and Versatile Evaluation Method of Thermal Stability of NI HTS Magnets
Authors: Mato, Takanobu; Noguchi, So
Abstract: In this article, a simple and versatile method of thermal stability evaluation for no-insulation (NI) high-temperature superconducting (HTS) magnets is proposed. Thermal stability is fundamental to superconducting magnets. The evaluation of coil temperature is an essential part of magnet design even though NI HTS magnets exhibit high thermal stability. Already proposed methods with complicated equivalent circuits, such as a partial element equivalent circuit (PEEC) model and a network model, require a long computation time and complication in coding. Hence, a simple way to evaluate the thermal stability of NI HTS coils is strongly desired; e.g., for a preliminary-design purpose. A method proposed in the article is a simple analytical formula derived from an equivalent RL-parallel circuit model of an NI HTS coil. The formulation considers a cooling effect with a simple assumption and Joule heating on radial (turn-to-turn contact) resistances. For a trial of the proposed simple evaluation, the thermal stability investigation is presented comparing the temperatures of the sudden discharge and ramp down. The results of the proposed analytical method are also compared with the PEEC model.2023-12-26T15:00:00ZMato, TakanobuNoguchi, SoIn this article, a simple and versatile method of thermal stability evaluation for no-insulation (NI) high-temperature superconducting (HTS) magnets is proposed. Thermal stability is fundamental to superconducting magnets. The evaluation of coil temperature is an essential part of magnet design even though NI HTS magnets exhibit high thermal stability. Already proposed methods with complicated equivalent circuits, such as a partial element equivalent circuit (PEEC) model and a network model, require a long computation time and complication in coding. Hence, a simple way to evaluate the thermal stability of NI HTS coils is strongly desired; e.g., for a preliminary-design purpose. A method proposed in the article is a simple analytical formula derived from an equivalent RL-parallel circuit model of an NI HTS coil. The formulation considers a cooling effect with a simple assumption and Joule heating on radial (turn-to-turn contact) resistances. For a trial of the proposed simple evaluation, the thermal stability investigation is presented comparing the temperatures of the sudden discharge and ramp down. The results of the proposed analytical method are also compared with the PEEC model.Multi-objective optimization of permanent magnet motors using deep learning and CMA-ES
http://hdl.handle.net/2115/91227
Title: Multi-objective optimization of permanent magnet motors using deep learning and CMA-ES
Authors: Mikami, Ryosuke; Sato, Hayaho; Hayashi, Shogo; Igarashi, Hajime
Abstract: This paper proposes a multi-objective optimization method for permanent magnet motors using a fast optimization algorithm, Covariance Matrix Adaptation Evolution Strategy (CMA-ES), and deep learning. Multi-objective optimization with topology optimization is effective in the design of permanent magnet motors. Although CMA-ES needs fewer population size than genetic algorithm for single objective problems, this is not evident for multi-objective problems. For this reason, the proposed method generates training data by solving the single-objective optimization multiple times using CMA-ES, and constructs a deep neural network (NN) based on the data to predict performance from motor images at high speed. The deep NN is then used for fast solution of multi-objective optimization problems. Numerical examples demonstrate the effectiveness of the proposed method.2023-12-13T15:00:00ZMikami, RyosukeSato, HayahoHayashi, ShogoIgarashi, HajimeThis paper proposes a multi-objective optimization method for permanent magnet motors using a fast optimization algorithm, Covariance Matrix Adaptation Evolution Strategy (CMA-ES), and deep learning. Multi-objective optimization with topology optimization is effective in the design of permanent magnet motors. Although CMA-ES needs fewer population size than genetic algorithm for single objective problems, this is not evident for multi-objective problems. For this reason, the proposed method generates training data by solving the single-objective optimization multiple times using CMA-ES, and constructs a deep neural network (NN) based on the data to predict performance from motor images at high speed. The deep NN is then used for fast solution of multi-objective optimization problems. Numerical examples demonstrate the effectiveness of the proposed method.Deep Neural Networks Based End-to-End DOA Estimation System
http://hdl.handle.net/2115/91226
Title: Deep Neural Networks Based End-to-End DOA Estimation System
Authors: Ando, Daniel Akira; Kase, Yuya; Nishimura, Toshihiko; Sato, Takanori; Ohganey, Takeo; Ogawa, Yasutaka; Hagiwara, Junichiro
Abstract: Direction of arrival (DOA) estimation is an antenna array signal processing technique used in, for instance, radar and sonar systems, source localization, and channel state information retrieval. As new applications and use cases appear with the development of next generation mobile communications systems, DOA estimation performance must be continually increased in order to support the nonstop growing demand for wireless technologies. In previous works, we verified that a deep neural network (DNN) trained offline is a strong candidate tool with the promise of achieving great on-grid DOA estimation performance, even compared to traditional algorithms. In this paper, we propose new techniques for further DOA estimation accuracy enhancement incorporating signal-to-noise ratio (SNR) prediction and an end-to-end DOA estimation system, which consists of three components: source number estimator, DOA angular spectrum grid estimator, and DOA detector. Here, we expand the performance of the DOA detector and angular spectrum estimator, and present a new solution for source number estimation based on DNN with very simple design. The proposed DNN system applied with said enhancement techniques has shown great estimation performance regarding the success rate metric for the case of two radio wave sources although not fully satisfactory results are obtained for the case of three sources.2023-11-30T15:00:00ZAndo, Daniel AkiraKase, YuyaNishimura, ToshihikoSato, TakanoriOhganey, TakeoOgawa, YasutakaHagiwara, JunichiroDirection of arrival (DOA) estimation is an antenna array signal processing technique used in, for instance, radar and sonar systems, source localization, and channel state information retrieval. As new applications and use cases appear with the development of next generation mobile communications systems, DOA estimation performance must be continually increased in order to support the nonstop growing demand for wireless technologies. In previous works, we verified that a deep neural network (DNN) trained offline is a strong candidate tool with the promise of achieving great on-grid DOA estimation performance, even compared to traditional algorithms. In this paper, we propose new techniques for further DOA estimation accuracy enhancement incorporating signal-to-noise ratio (SNR) prediction and an end-to-end DOA estimation system, which consists of three components: source number estimator, DOA angular spectrum grid estimator, and DOA detector. Here, we expand the performance of the DOA detector and angular spectrum estimator, and present a new solution for source number estimation based on DNN with very simple design. The proposed DNN system applied with said enhancement techniques has shown great estimation performance regarding the success rate metric for the case of two radio wave sources although not fully satisfactory results are obtained for the case of three sources.An Explainable Recommendation Based on Acyclic Paths in an Edge-Colored Graph
http://hdl.handle.net/2115/91071
Title: An Explainable Recommendation Based on Acyclic Paths in an Edge-Colored Graph
Authors: Chinone, Kosuke; Nakamura, Atsuyoshi
Abstract: We propose a novel recommendation algorithm based on acyclic paths in an edge-colored graph. In our method, all the objects including users, items to recommend, and other things usable to recommendation are represented as vertices in an edge-colored directed graph, in which edge color represents relation between the objects of its both ends. By setting each edge weight appropriately so as to reflect how much the object corresponding to its one end is preferred by people who prefer the object corresponding to its other end, the probability of an s-t path, which is defined as the product of its component edges' weights, can be regarded as preference degree of item t (item corresponding to vertex t) by user s (user corresponding to vertex s) in the context represented by the path. Given probability threshold θ, the proposed algorithm recommends user s to item t that has high sum of the probabilities of all the acyclic s-t paths whose probability is at least θ. For item t recommended to user s, the algorithm also shows high probability color sequences of those s-t paths, from which we can know main contexts of the recommendation of item t for user s. According to our experiments using real-world datasets, the recommendation performance of our method is comparable to the non-explainable state-of-the-art recommendation methods.2022-03-18T15:00:00ZChinone, KosukeNakamura, AtsuyoshiWe propose a novel recommendation algorithm based on acyclic paths in an edge-colored graph. In our method, all the objects including users, items to recommend, and other things usable to recommendation are represented as vertices in an edge-colored directed graph, in which edge color represents relation between the objects of its both ends. By setting each edge weight appropriately so as to reflect how much the object corresponding to its one end is preferred by people who prefer the object corresponding to its other end, the probability of an s-t path, which is defined as the product of its component edges' weights, can be regarded as preference degree of item t (item corresponding to vertex t) by user s (user corresponding to vertex s) in the context represented by the path. Given probability threshold θ, the proposed algorithm recommends user s to item t that has high sum of the probabilities of all the acyclic s-t paths whose probability is at least θ. For item t recommended to user s, the algorithm also shows high probability color sequences of those s-t paths, from which we can know main contexts of the recommendation of item t for user s. According to our experiments using real-world datasets, the recommendation performance of our method is comparable to the non-explainable state-of-the-art recommendation methods.Learning shared embedding representation of motion and text using contrastive learning
http://hdl.handle.net/2115/91020
Title: Learning shared embedding representation of motion and text using contrastive learning
Authors: Horie, Junpei; Noguchi, Wataru; Iizuka, Hiroyuki; Yamamoto, Masahito
Abstract: Multimodal learning of motion and text tries to find the correspondence between skeletal time-series data acquired by motion capture and the text that describes the motion. In this field, good associations can realize both motion-to-text and text-to-motion applications. However, the previous methods failed to associate motion with text, taking into account details of descriptions, for example, whether to move the left or right arm. In this paper, we propose a motion-text contrastive learning method for making correspondences between motion and text in a shared embedding space. We showed that our model outperforms the previous studies in the task of action recognition. We also qualitatively show that, by using a pre-trained text encoder, our model can perform motion retrieval with detailed correspondences between motion and text.2022-12-26T15:00:00ZHorie, JunpeiNoguchi, WataruIizuka, HiroyukiYamamoto, MasahitoMultimodal learning of motion and text tries to find the correspondence between skeletal time-series data acquired by motion capture and the text that describes the motion. In this field, good associations can realize both motion-to-text and text-to-motion applications. However, the previous methods failed to associate motion with text, taking into account details of descriptions, for example, whether to move the left or right arm. In this paper, we propose a motion-text contrastive learning method for making correspondences between motion and text in a shared embedding space. We showed that our model outperforms the previous studies in the task of action recognition. We also qualitatively show that, by using a pre-trained text encoder, our model can perform motion retrieval with detailed correspondences between motion and text.COVID-19 detection based on self-supervised transfer learning using chest X-ray images
http://hdl.handle.net/2115/90979
Title: COVID-19 detection based on self-supervised transfer learning using chest X-ray images
Authors: Li, Guang; Togo, Ren; Ogawa, Takahiro; Haseyama, Miki
Abstract: Purpose
Considering several patients screened due to COVID-19 pandemic, computer-aided detection has strong potential in assisting clinical workflow efficiency and reducing the incidence of infections among radiologists and healthcare providers. Since many confirmed COVID-19 cases present radiological findings of pneumonia, radiologic examinations can be useful for fast detection. Therefore, chest radiography can be used to fast screen COVID-19 during the patient triage, thereby determining the priority of patient’s care to help saturated medical facilities in a pandemic situation.
Methods
In this paper, we propose a new learning scheme called self-supervised transfer learning for detecting COVID-19 from chest X-ray (CXR) images. We compared six self-supervised learning (SSL) methods (Cross, BYOL, SimSiam, SimCLR, PIRL-jigsaw, and PIRL-rotation) with the proposed method. Additionally, we compared six pretrained DCNNs (ResNet18, ResNet50, ResNet101, CheXNet, DenseNet201, and InceptionV3) with the proposed method. We provide quantitative evaluation on the largest open COVID-19 CXR dataset and qualitative results for visual inspection.
Results
Our method achieved a harmonic mean (HM) score of 0.985, AUC of 0.999, and four-class accuracy of 0.953. We also used the visualization technique Grad-CAM++ to generate visual explanations of different classes of CXR images with the proposed method to increase the interpretability.
Conclusions
Our method shows that the knowledge learned from natural images using transfer learning is beneficial for SSL of the CXR images and boosts the performance of representation learning for COVID-19 detection. Our method promises to reduce the incidence of infections among radiologists and healthcare providers.2023-03-31T15:00:00ZLi, GuangTogo, RenOgawa, TakahiroHaseyama, MikiPurpose
Considering several patients screened due to COVID-19 pandemic, computer-aided detection has strong potential in assisting clinical workflow efficiency and reducing the incidence of infections among radiologists and healthcare providers. Since many confirmed COVID-19 cases present radiological findings of pneumonia, radiologic examinations can be useful for fast detection. Therefore, chest radiography can be used to fast screen COVID-19 during the patient triage, thereby determining the priority of patient’s care to help saturated medical facilities in a pandemic situation.
Methods
In this paper, we propose a new learning scheme called self-supervised transfer learning for detecting COVID-19 from chest X-ray (CXR) images. We compared six self-supervised learning (SSL) methods (Cross, BYOL, SimSiam, SimCLR, PIRL-jigsaw, and PIRL-rotation) with the proposed method. Additionally, we compared six pretrained DCNNs (ResNet18, ResNet50, ResNet101, CheXNet, DenseNet201, and InceptionV3) with the proposed method. We provide quantitative evaluation on the largest open COVID-19 CXR dataset and qualitative results for visual inspection.
Results
Our method achieved a harmonic mean (HM) score of 0.985, AUC of 0.999, and four-class accuracy of 0.953. We also used the visualization technique Grad-CAM++ to generate visual explanations of different classes of CXR images with the proposed method to increase the interpretability.
Conclusions
Our method shows that the knowledge learned from natural images using transfer learning is beneficial for SSL of the CXR images and boosts the performance of representation learning for COVID-19 detection. Our method promises to reduce the incidence of infections among radiologists and healthcare providers.Optimal reconstruction of noisy dynamics and selection probabilities in Boolean networks
http://hdl.handle.net/2115/90943
Title: Optimal reconstruction of noisy dynamics and selection probabilities in Boolean networks
Authors: Kobayashi, Koichi; Wu, Yuhu
Abstract: In the analysis and control of complex systems, including gene regulatory networks, it is important to reconstruct a mathematical model from a priori information and noisy experimental data. A Boolean network (BN) is well known as a mathematical model of gene regulatory networks. Each state of BNs takes a binary value (0 or 1), and its update rule is given by a set of Boolean functions. In this paper, we consider the optimal reconstruction problem of finding a probabilistic BN consisting of the main dynamics and the noisy dynamics, by giving the main dynamics and the sample mean of the state obtained from noisy experimental data. In the proposed method, the selection probability of the main dynamics is maximized. We show that the optimal Boolean function of the noisy dynamics is a constant (0 or 1) map under no assumption on the structure of noisy dynamics. Finally, as a biological application, the reconstruction of a PBN model of the lac operon networks of Escherichia coli bacterium is addressed using the proposed approach.2022-01-31T15:00:00ZKobayashi, KoichiWu, YuhuIn the analysis and control of complex systems, including gene regulatory networks, it is important to reconstruct a mathematical model from a priori information and noisy experimental data. A Boolean network (BN) is well known as a mathematical model of gene regulatory networks. Each state of BNs takes a binary value (0 or 1), and its update rule is given by a set of Boolean functions. In this paper, we consider the optimal reconstruction problem of finding a probabilistic BN consisting of the main dynamics and the noisy dynamics, by giving the main dynamics and the sample mean of the state obtained from noisy experimental data. In the proposed method, the selection probability of the main dynamics is maximized. We show that the optimal Boolean function of the noisy dynamics is a constant (0 or 1) map under no assumption on the structure of noisy dynamics. Finally, as a biological application, the reconstruction of a PBN model of the lac operon networks of Escherichia coli bacterium is addressed using the proposed approach.Design of PLC E31-E13 and E-LP tapered mode converters using fast quasiadiabatic dynamics
http://hdl.handle.net/2115/90935
Title: Design of PLC E31-E13 and E-LP tapered mode converters using fast quasiadiabatic dynamics
Authors: Wang, Han; Fujisawa, Takeshi; Sato, Takanori; Wada, Masaki; Mori, Takayoshi; Sakamoto, Taiji; Imada, Ryota; Matsui, Takashi; Nakajima, Kazuhide; Saitoh, Kunimasa
Abstract: The previous PLC E-31 and E-13 mode converters employ linear tapered structures to fulfill the adiabatic coupling condition and achieve a high mode conversion efficiency, which requires long taper lengths. In this paper, we propose two types of PLC tapered structures that can be applied in different scenarios using fast quasiadiabatic dynamics. These structures enable mode conversions of E31(31)-E-13 modes, as well as E-31-LP0(2) and E-13- LP21b modes, respectively. The E-31-E-13 mode converter can be reduced from the previous three-stage linear taper of 12000 μm to approximately 4000 μm in length. Similarly, the E-31-LP02, E-13-LP21b mode converter can be reduced from the previous two-stage linear taper of over 6000 μm to approximately 4000 μm in length.2023-12-09T15:00:00ZWang, HanFujisawa, TakeshiSato, TakanoriWada, MasakiMori, TakayoshiSakamoto, TaijiImada, RyotaMatsui, TakashiNakajima, KazuhideSaitoh, KunimasaThe previous PLC E-31 and E-13 mode converters employ linear tapered structures to fulfill the adiabatic coupling condition and achieve a high mode conversion efficiency, which requires long taper lengths. In this paper, we propose two types of PLC tapered structures that can be applied in different scenarios using fast quasiadiabatic dynamics. These structures enable mode conversions of E31(31)-E-13 modes, as well as E-31-LP0(2) and E-13- LP21b modes, respectively. The E-31-E-13 mode converter can be reduced from the previous three-stage linear taper of 12000 μm to approximately 4000 μm in length. Similarly, the E-31-LP02, E-13-LP21b mode converter can be reduced from the previous two-stage linear taper of over 6000 μm to approximately 4000 μm in length.Compressed gastric image generation based on soft-label dataset distillation for medical data sharing
http://hdl.handle.net/2115/90758
Title: Compressed gastric image generation based on soft-label dataset distillation for medical data sharing
Authors: Li, Guang; Togo, Ren; Ogawa, Takahiro; Haseyama, Miki
Abstract: Background and objective: Sharing of medical data is required to enable the cross-agency flow of health-care information and construct high-accuracy computer-aided diagnosis systems. However, the large sizes of medical datasets, the massive amount of memory of saved deep convolutional neural network (DCNN) models, and patients ' privacy protection are problems that can lead to inefficient medical data sharing. Therefore, this study proposes a novel soft-label dataset distillation method for medical data sharing. Methods: The proposed method distills valid information of medical image data and generates several compressed images with different data distributions for anonymous medical data sharing. Furthermore, our method can extract essential weights of DCNN models to reduce the memory required to save trained models for efficient medical data sharing. Results: The proposed method can compress tens of thousands of images into several soft-label images and reduce the size of a trained model to a few hundredths of its original size. The compressed images obtained after distillation have been visually anonymized; therefore, they do not contain the private in-formation of the patients. Furthermore, we can realize high-detection performance with a small number of compressed images. Conclusions: The experimental results show that the proposed method can improve the efficiency and security of medical data sharing.2022-11-30T15:00:00ZLi, GuangTogo, RenOgawa, TakahiroHaseyama, MikiBackground and objective: Sharing of medical data is required to enable the cross-agency flow of health-care information and construct high-accuracy computer-aided diagnosis systems. However, the large sizes of medical datasets, the massive amount of memory of saved deep convolutional neural network (DCNN) models, and patients ' privacy protection are problems that can lead to inefficient medical data sharing. Therefore, this study proposes a novel soft-label dataset distillation method for medical data sharing. Methods: The proposed method distills valid information of medical image data and generates several compressed images with different data distributions for anonymous medical data sharing. Furthermore, our method can extract essential weights of DCNN models to reduce the memory required to save trained models for efficient medical data sharing. Results: The proposed method can compress tens of thousands of images into several soft-label images and reduce the size of a trained model to a few hundredths of its original size. The compressed images obtained after distillation have been visually anonymized; therefore, they do not contain the private in-formation of the patients. Furthermore, we can realize high-detection performance with a small number of compressed images. Conclusions: The experimental results show that the proposed method can improve the efficiency and security of medical data sharing.Sample Complexity of Learning Multi-value Opinions in Social Networks
http://hdl.handle.net/2115/90677
Title: Sample Complexity of Learning Multi-value Opinions in Social Networks
Authors: Shinoda, Masato; Sakurai, Yuko; Oyama, Satoshi
Abstract: We consider how many users we need to query in order to estimate the extent to which multi-value opinions (information) have propagated in a social network. For example, if the launch date of a new product has changed many times, the company might want to know to which people the most current information has reached. In the propagation model we consider, the social network is represented as a directed graph, and an agent (node) updates its state if it receives a stronger opinion (updated information) and then forwards the opinion in accordance with the direction of its edges. Previous work evaluated opinion propagation in a social network by using the probably approximately correct (PAC) learning framework and considered only binary opinions. In general, PAC learnability, i.e., the finiteness of the number of samples needed, is not guaranteed when generalizing from a binary-value model to a multi-value model. We show that the PAC learnability of multi-value opinions propagating in a social network. We first prove that the number of samples needed in a multi-opinion model is sufficient for (k−1)log(k−1) times the number of samples needed in a binary-opinion model, when k (≥3) is the number of opinions. We next prove that the upper and lower bounds on the number of samples needed to learn a multi-opinion model can be determined from the Natarajan dimension, which is a generalization of the Vapnik-Chervonenkis dimension.2022-11-11T15:00:00ZShinoda, MasatoSakurai, YukoOyama, SatoshiWe consider how many users we need to query in order to estimate the extent to which multi-value opinions (information) have propagated in a social network. For example, if the launch date of a new product has changed many times, the company might want to know to which people the most current information has reached. In the propagation model we consider, the social network is represented as a directed graph, and an agent (node) updates its state if it receives a stronger opinion (updated information) and then forwards the opinion in accordance with the direction of its edges. Previous work evaluated opinion propagation in a social network by using the probably approximately correct (PAC) learning framework and considered only binary opinions. In general, PAC learnability, i.e., the finiteness of the number of samples needed, is not guaranteed when generalizing from a binary-value model to a multi-value model. We show that the PAC learnability of multi-value opinions propagating in a social network. We first prove that the number of samples needed in a multi-opinion model is sufficient for (k−1)log(k−1) times the number of samples needed in a binary-opinion model, when k (≥3) is the number of opinions. We next prove that the upper and lower bounds on the number of samples needed to learn a multi-opinion model can be determined from the Natarajan dimension, which is a generalization of the Vapnik-Chervonenkis dimension.