检索结果-内蒙古大学图书馆

38th International Conference on High Performance Computing, ISC High Performance 2023

作者： Boku, Taisuke Tsunashima, Ryuta Kobayashi, Ryohei Fujita, Norihisa Lee, Seyong Vetter, Jeffrey S. Murai, Hitoshi Nakao, Masahiro Tsuji, Miwako Sato, Mitsuhisa Center for Computational Sciences University of Tsukuba Tsukuba Japan Degree Programs in Systems and Information Engineering University of Tsukuba Tsukuba Japan Oak Ridge National Laboratory Oak Ridge United States RIKEN Center for Computational Science Kobe Japan

ISBN: (纸本)9783031408427

Accelerated computing in HPC such as with GPU, plays a central role in HPC nowadays. However, in some complicated applications with partially different performance behavior is hard to solve with a single type of accelerator where GPU is not the perfect solution in these cases. We are developing a framework and transpiler allowing the users to program the codes with a single notation of OpenACC to be compiled for multi-hybrid accelerators, named MHOAT (Multi-Hybrid OpenACC Translator) for HPC applications. MHOAT parses the original code with directives to identify the target accelerating devices, currently supporting NVIDIA GPU and Intel FPGA, dispatching these specific partial codes to background compilers such as NVIDIA HPC SDK for GPU and OpenARC research compiler for FPGA, then assembles binaries for the final object with FPGA bitstream file. In this paper, we present the concept, design, implementation, and performance evaluation of a practical astrophysics simulation code where we successfully enhanced the performance up to 10 times faster than the GPU-only solution. © 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.

关键词： Field programmable gate arrays (FPGA)

来源：评论

学校读者我要写书评

暂无评论

CAMRI Loss: Improving the Recall of a Specific Class without Sacrificing Accuracy

CAMRI Loss: Improving the Recall of a Specific Class without...

引用

作者： Nishiyama, Daiki Fukuchi, Kazuto Akimoto, Youhei Sakuma, Jun Master's Program in Computer Science Degree Programs in Systems and Information Engineering Graduate School of Science and Technology University of Tsukuba Tsukuba-shi305-8577 Japan RIKEN Center for Advanced Intelligence Project Tokyo103-0027 Japan Department of Computer Science Faculty of Engineering Information and Systems University of Tsukuba Tsukuba-shi305-8577 Japan

In real world applications of multiclass classification models, misclassification in an important class (e.g., stop sign) can be significantly more harmful than in other classes (e.g., no parking). Thus, it is crucial to improve the recall of an important class while maintaining overall accuracy. For this problem, we found that improving the separation of important classes relative to other classes in the feature space is effective. Existing methods that give a class-sensitive penalty for cross-entropy loss do not improve the separation. Moreover, the methods designed to improve separations between all classes are unsuitable for our purpose because they do not consider the important classes. To achieve the separation, we propose a loss function that explicitly gives loss for the feature space, called class-sensitive additive angular margin (CAMRI) loss. CAMRI loss is expected to reduce the variance of an important class due to the addition of a penalty to the angle between the important class features and the corresponding weight vectors in the feature space. In addition, concentrating the penalty on only the important class hardly sacrifices separating the other classes. Experiments on CIFAR-10, GTSRB, and AwA2 showed that CAMRI loss could improve the recall of a specific class without sacrificing accuracy. In particular, compared with GTSRB's second-worst class recall when trained with cross-entropy loss, CAMRI loss improved recall by 9%. © 2023 The Institute of Electronics, information and Communication Engineers.

关键词： Deep learning

来源：评论

学校读者我要写书评

暂无评论

CNN Model Compression by Merit-Based Distillation

CNN Model Compression by Merit-Based Distillation

引用

International Colloquium on Signal Processing & Its Applications, CSPA

作者： Takumi Morikawa Keisuke Kameyama Degree Programs in Systems and Information Engineering Graduate School of Science and Technology University of Tsukuba Tsukuba Japan Faculty of Engineering Information and Systems University of Tsukuba Tsukuba Japan

Deep learning has advanced dramatically in recent years, and especially large convolutional neural networks (CNNs) have shown outstanding performance in a wide variety of tasks. However, such large-scale CNNs may not be able to be utilized in practice because they require a lot of computational resources. Therefore, there is a need to realize a CNN in a small model that has high performance similar to the large CNN. One method to solve these problems is Distillation, which uses a model of large scale and high performance as the teacher and a small model as the student model, and performs model compression by transferring the knowledge of the teacher model to the student model. However, in distillation, even when the teacher model incorrectly guesses the training input, the student model learns to mimic the teacher. This can prevent the student model from learning better. In this paper, we propose Meritbased Distillation, in which the accuracy of the teacher model’s output determines the degree to which the student model mimics the teacher model’s output, thereby preventing the student to learn undesired mappings. In the experiment, the effect of the Merit-based Distillation is evaluated. Consequently, the proposed method outperforms the improvements in the generalization by the student models trained by other known methods.

关键词： Training Deep learning Computational modeling Convolutional neural networks Task analysis

来源：评论

学校读者我要写书评

暂无评论

Informative Band Subset Selection for Hyperspectral Image Classification using Joint and Conditional Mutual information

Informative Band Subset Selection for Hyperspectral Image Cl...

引用

IEEE Symposium Series on Computational Intelligence (SSCI)

作者： U A Md Ehsan Ali Keisuke Kameyama Degree Programs in Systems and Information Engineering Graduate School of Science and Technology University of Tsukuba Tsukuba Japan Faculty of Engineering Information and Systems University of Tsukuba Tsukuba Japan

ISBN: (纸本)9781665487696

Hundreds of contiguous bands of remotely sensed hyperspectral image (HSI) capture the spectral signatures of observed objects or materials on the earth's surface. Although the HSI data is able to provide huge information with great details, it poses challenges to image analysis because of the high computational cost due to the large dimensionality of the feature space, redundancy in information, and curse of dimensionality. To overcome these difficulties, this paper proposes an information-theoretic feature selection approach for selecting an informative feature subset considering the maximum of the minimum approach. The minimum of conditional mutual information based estimation is used to select a feature among the selected feature subset. This selected feature with the corresponding candidate feature is then exploited using the mutual information and joint mutual information based valuation to find the maximum relevance of the candidate features with the target classes. The effectiveness of the proposed approach, called joint-conditional mutual information for selecting informative feature (JCIF), is assessed by implementing it in some synthetic data and three remotely sensed HSI data. Several known feature selection algorithms are also used for comparison purposes. The results of the experiments with K-Nearest Neighbors and Support Vector Machine classifiers reveal that JCIF performs better in selecting informative features.

关键词： Support vector machines Reflectivity Earth Redundancy Estimation Feature extraction Mutual information

来源：评论

学校读者我要写书评

暂无评论

Machine Learning Curriculums Generated by Classifier Ensembles

Machine Learning Curriculums Generated by Classifier Ensembl...

引用

International Colloquium on Signal Processing & Its Applications, CSPA

作者： Tzu-Jui Huang Keisuke Kameyama Degree Programs in Systems and Information Engineering Graduate School of Science and Technology University of Tsukuba Tsukuba Japan Faculty of Engineering Information and Systems University of Tsukuba Tsukuba Japan

Image classification is a fundamental task that attempts to classify the images into the classes they belong to. CNN is commonly used for the classification due to its flexibility and stability. This paper treats Curriculum Learning as an assistant method that decides the procedure the CNN ought to be trained. Curriculum Learning is generally defined as a method that divides and orders the training datasets into several different levels of learning difficulties. This paper utilizes the idea of ensemble classification that trains according to the order based on the consensus of different machine learning models. As a result, the proposed Ensemble-Generated Curriculum Learning model achieves higher accuracy compared to that of a single-stage CNN learning with dog face recognition and CIFAR-10 datasets.

关键词： Training Face recognition Stability criteria Machine learning Dogs Classification algorithms Task analysis

来源：评论

学校读者我要写书评

暂无评论

LSTM-based Forecasting using Policy Stringency and Time-varying Parameters of the SIR Model for COVID-19

LSTM-based Forecasting using Policy Stringency and Time-vary...

引用

International Colloquium on Signal Processing & Its Applications, CSPA

作者： Pavodi Maniamfu Keisuke Kameyama Degree Programs in Systems and Information Engineering Graduate School of Science and Technology University of Tsukuba Tsukuba Japan Faculty of Engineering Information and Systems University of Tsukuba Tsukuba Japan

Accurate forecasting of the number of infections is an important task that can allow health care decision makers to allocate medical resources efficiently during a pandemic. Two approaches have been combined, a stochastic model by Vega et al. for modelling infectious disease and Long Short-Term Memory using COVID-19 data and government’s policies. In the proposed model, LSTM functions as a nonlinear adaptive filter to modify the outputs of the SIR model for more accurate forecasts one to four weeks in the future. Our model outperforms most models among the CDC models using the United States data. We also applied the model on the Canadian data from two provinces, Saskatchewan and Ontario where it performs with a low mean absolute percentage error.

关键词： COVID-19 Adaptation models Pandemics Infectious diseases Stochastic processes Predictive models Data models

来源：评论

学校读者我要写书评

暂无评论

Study of Galaxy Collisions and Thermodynamic Evolution of Gas Using the Exact Integration Scheme 22nd

Study of Galaxy Collisions and Thermodynamic Evolution of...

引用

22nd International Conference on Computational Science and Its Applications , ICCSA 2022

作者： Otaki, Koki Mori, Masao Degree Programs in Pure and Applied Sciences Graduate School of Science and Technology University of Tsukuba Tennodai 1-1-1 Ibaraki Tsukuba305-8577 Japan Degree Programs in Systems and Information Engineering Graduate School of Science and Technology University of Tsukuba Tennodai 1-1-1 Ibaraki Tsukuba305-8577 Japan Center for Computational Sciences University of Tsukuba Tennodai 1-1-1 Ibaraki Tsukuba305-8577 Japan

ISBN: (纸本)9783031105616

Radiative cooling of the interstellar medium plays a vital role in the context of galaxy formation and evolution. On the other hand, the cooling time in the high-density regions involving star formation is much shorter than the dynamical time of the gas. In numerical simulations, it is challenging to solve physical phenomena coexisting on significantly different timescales, and it is known as the overcooling problem in the study of galaxy formation. Townsend (2009) has developed the Exact Integration (EI) scheme that provides a stable solution for the cooling term in the energy equation of astrophysical fluid dynamics, regardless of the size of the simulation time step. We apply the EI scheme to define the effective cooling time that accounts for the temperature dependence of the cooling rate and investigate the thermodynamic evolution of gas in colliding dark matter subhalos. The results show that the conventional cooling time always indicates a shorter than the effective cooling time derived by the EI scheme because it does not include the dependence of the cooling rate on temperature. Furthermore, we run three-dimensional galaxy collision simulations to examine the difference in thermodynamic evolution between the EI scheme and the conventional Crank–Nicholson method for solving the cooling equation. Comparing the results of the two simulations, we find that the EI scheme suppresses the rapid temperature decrease after galaxy collisions. Thus, the EI scheme indicates considerable potential for solving the overcooling problem in the study of galaxy formation. © 2022, The Author(s).

关键词： Galaxies

来源：评论

学校读者我要写书评

暂无评论

Variable-Speed Teaching-Playback as Real-World Data Augmentation for Imitation Learning

arXiv

引用

arXiv 2024年

作者： Masuya, Nozomu Sato, Hiroshi Yamane, Koki Kusume, Takuya Sakaino, Sho Tsuji, Toshiaki Master's Program in Intelligent and Mechanical Interaction Systems Degree Programs in Systems and Information and Engineering Graduate School of Science and Technology University of Tsukuba Japan Doctoral Program in Intelligent and Mechanical Interaction Systems Degree Programs in Systems and Information and Engineering Graduate School of Science and Technology University of Tsukuba Japan Department of Intelligent Interaction Technologies Institute of Systems and Information Engineering University of Tsukuba Japan Graduate School of Science and Engineering Saitama University Saitama Japan

Because imitation learning relies on human demonstrations in hard-to-simulate settings, the inclusion of force control in this method has resulted in a shortage of training data, even with a simple change in speed. Although the field of data augmentation has addressed the lack of data, conventional methods of data augmentation for robot manipulation are limited to simulation-based methods or downsampling for position control. This paper proposes a novel method of data augmentation that is applicable to force control and preserves the advantages of real-world datasets. We applied teaching-playback at variable speeds as real-world data augmentation to increase both the quantity and quality of environmental reactions at variable speeds. An experiment was conducted on bilateral control-based imitation learning using a method of imitation learning equipped with position-force control. We evaluated the effect of real-world data augmentation on two tasks, pick-and-place and wiping, at variable speeds, each from two human demonstrations at fixed speed. The results showed a maximum 55% increase in success rate from a simple change in speed of real-world reactions and improved accuracy along the duration/frequency command by gathering environmental reactions at variable speeds. Copyright © 2024, The Authors. All rights reserved.

关键词： Demonstrations

来源：评论

学校读者我要写书评

暂无评论

On a New Rough Clustering Based on Optimization

On a New Rough Clustering Based on Optimization

引用

International Symposium on Soft Computing and Intelligent systems (SCIS)

作者： Sakuya FURUTA Yasunori ENDO Master's Program in Risk and Resilience Engineering Degree Programs in Systems and Information Engineering Graduate School of Science annd Technology University of Tsukuba Ibaraki Japan Institute of Systems and Information Engineering University of Tsukuba Ibaraki Japan

ISBN: (数字)9798350373332

ISBN: (纸本)9798350373349

The purpose of this study is to develop and investigate a new rough clustering based on optimization. We propose a restructured algorithm using only one variable, instead of the algorithm described using two variables in existing methods. We also perform numerical examples using artificial data and the Iris and Wine datasets. This research has the potential to be used as a method for determining false negatives and false positives in cancer diagnosis.

关键词： Iris Accuracy Clustering algorithms Computational complexity Optimization Cancer

来源：评论

学校读者我要写书评

暂无评论

THE APPLICATION OF ARTIFICIAL-INTELLIGENCE TO FUTURE TACTICAL C2 DESIGN

引用

NAVAL ENGINEERS JOURNAL 1985年第4期97卷 287-295页

作者： MURRAY, AJ The Author is a senior analyst in the Major Programs Group at TITAN Systems Inc. Vienna Virginia. His background and experience in both the private and public sectors covers ten years of systems engineering and management for tactical and strategic C3 weapons systems and information processing systems. Recently Mr. Murray developed a preliminary C3architecture to support strategic air defense and participated in the development of a C3architecture for PACOM. Mr. Murray holds a B.S.E.E. degree from Lehigh University and an M.E.A. degree from George Washington University where he is currently a doctoral candidate specializing in the field of artificial intelligence. His memberships include ASNE and the United States Naval Institute.

The purpose of this paper is to develop a top-level design for an artificial intelligence based C 2 decision support system. The design objective is to provide a command level decision maker with timely and accurate information to support optimal utilization of assets to be employed in theater tactical combat situations. This paper will be divided into two portions. The first will discuss artificial intelligence (AI) in the general sense, i.e., what AI is and what it is not , and why such an approach could have broad-sweeping applications in naval battle command and control. A generic AI-based decision support system will be illustrated. This system will have three major attributes: 1) intelligent interfaces to the operator, 2) a goal-seeking problem processing module, and 3) a knowledge base which utilizes interacting intelligent data base and model base elements. It will be shown how all three of these attributes can be effectively combined to provide the force commander with a rapid assessment of a given threat situation and aid in the selection of recommended courses of action. The second part of this paper will be directed toward examining specific applications of decision support system technology, using battle management and force management as exemplary domains. Key input drivers, decision algorithms and system safeguards will be addressed. Comparisons of predicted DSS performance vs. current architectures will be made, and discussions of interfacing to existing networks will also be included.

关键词： NAVAL WARFARE

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：