检索结果-内蒙古大学图书馆

arXiv 2023年

作者： Ren, Xiaozhe Zhou, Pingyi Meng, Xinfan Huang, Xinjing Wang, Yadao Wang, Weichao Li, Pengfei Zhang, Xiaoda Podolskiy, Alexander Arshinov, Grigory Bout, Andrey Piontkovskaya, Irina Wei, Jiansheng Jiang, Xin Su, Teng Liu, Qun Yao, Jun Noah’s Ark Lab Huawei Technologies China Distributed and Parallel Software Lab Huawei Technologies China

The scaling of large language models has greatly improved natural language understanding, generation, and reasoning. In this work, we develop a system that trained a trillion-parameter language model on a cluster of Ascend 910 AI processors 2 and MindSpore framework 3, and present the language model with 1.085T parameters named PanGu-Σ. With parameter inherent from PanGu-α [1], we extend the dense Transformer model to sparse one with Random Routed Experts (RRE), and efficiently train the model over 329B tokens by using Expert Computation and Storage Separation (ECSS). This resulted in a 6.3x increase in training throughput through heterogeneous computing. Our experimental findings show that PanGu-Σ provides state-of-the-art performance in zero-shot learning of various Chinese NLP downstream tasks. Moreover, it demonstrates strong abilities when fine-tuned in application data of open-domain dialogue, question answering, machine translation and code generation. Copyright © 2023, The Authors. All rights reserved.

关键词： Zero-shot learning

来源：评论

学校读者我要写书评

暂无评论

LKFlowNet: A deep neural network based on large kernel convolution for fast and accurate nonlinear fluid-changing prediction

引用

PHYSICS OF FLUIDS 2024年第9期36卷

作者： Liu, Yan Zhang, Qingyang Chen, Xinhai Xu, Chuanfu Wang, Qinglin Liu, Jie Natl Univ Def Technol Lab Digitizing Software Frontier Equipment Changsha Peoples R China Natl Univ Def Technol Sci & Technol Parallel & Distributed Proc Lab Changsha Peoples R China

The rapid development of artificial intelligence has promoted the emergence of new flow field prediction methods. These methods address challenges posed by nonlinear problems and significantly reduce computational time and cost compared to traditional numerical simulations. However, they often struggle to capture the dynamic sparse characteristics of the flow field effectively. To bridge this gap, we introduce LKFlowNet, a new large kernel convolutional neural network specifically designed for complex flow fields in nonlinear fluid dynamics systems. LKFlowNet adopts a multi-branch large kernel convolution computing architecture, which can skillfully handle the complex nonlinear dynamic characteristics of flow changes. Drawing inspiration from the dilated convolution mechanism, we developed the RepDWConv block, a re-parameterized depthwise convolution that extends the convolutional kernel's coverage. This enhancement improves the model's ability to capture long-range dependencies and sparse structural features in fluid dynamics. Additionally, a customized physical loss function ensures accuracy and physical consistency in flow field reconstruction. Comparative studies reveal that LKFlowNet significantly outperforms existing neural network architectures, providing more accurate and physically consistent predictions in complex nonlinear variations such as velocity and pressure fields. The model demonstrates strong versatility and scalability, accurately predicting the flow field of various geometric configurations without modifying the architecture. This capability positions LKFlowNet as a promising new direction in fluid dynamics research, potentially revolutionizing flow field prediction by combining high efficiency and accuracy. Our results suggest that LKFlowNet could become an indispensable tool in intelligent flow field prediction, reshaping the analysis and processing of fluid dynamics.

关键词： Flow fields

来源：评论

学校读者我要写书评

暂无评论

IoUformer: Pseudo-IoU prediction with transformer for visual tracking

引用

NEURAL NETWORKS 2024年 170卷 548-563页

作者： Cai, Huayue Lan, Long Zhang, Jing Zhang, Xiang Zhan, Yibing Luo, Zhigang Natl Univ Def Technol Inst Quantum Changsha Peoples R China Natl Univ Def Technol State Key Lab High Performance Comp Changsha Peoples R China Natl Univ Def Technol Lab Digitizing Software Frontier Equipment Changsha Peoples R China Univ Sydney Sch Comp Sci Sydney Australia JD Explore Acad Beijing Peoples R China Natl Univ Def Technol Coll Comp Sci & Technol Parallel & Distributed Lab Changsha Peoples R China

Siamese tracking has witnessed tremendous progress in tracking paradigm. However, its default box estimation pipeline still faces a crucial inconsistency issue, namely, the bounding box decided by its classification score is not always best overlapped with the ground truth, thus harming performance. To this end, we explore a novel simple tracking paradigm based on the intersection over union (IoU) value prediction. To first bypass this inconsistency issue, we propose a concise target state predictor termed IoUformer, which instead of default box estimation pipeline directly predicts the IoU values related to tracking performance metrics. In detail, it extends the long-range dependency modeling ability of transformer to jointly grasp target-aware interactions between target template and search region, and search sub-region interactions, thus neatly unifying global semantic interaction and target state prediction. Thanks to this joint strength, IoUformer can predict reliable IoU values near-linear with the ground truth, which paves a safe way for our new IoU-based siamese tracking paradigm. Since it is non-trivial to explore this paradigm with pleased efficacy and portability, we offer the respective network components and two alternative localization ways. Experimental results show that our IoUformer-based tracker achieves promising results with less training data. For its applicability, it still serves as a refinement module to consistently boost existing advanced trackers.

关键词： Visual tracking Transformer Siamese network IoU prediction

来源：评论

学校读者我要写书评

暂无评论

A physics-informed neural network-based aerodynamic parameter identification method for aircraft

引用

PHYSICS OF FLUIDS 2025年第2期37卷

作者： Lin, Jie Chen, Shu-sheng Yang, Hua Jiang, Quan-feng Liu, Jie Natl Univ Def Technol Lab Digitizing Software Frontier Equipment Changsha 410073 Hunan Peoples R China Natl Univ Def Technol Natl Key Lab Parallel & Distributed Comp Changsha 410073 Hunan Peoples R China China Aerodynam Res & Dev Ctr Computat Aerodynam Inst Mianyang 621000 Sichuan Peoples R China Northwestern Polytech Univ Sch Aeronaut Xian 710072 Shaanxi Peoples R China Natl Key Lab Aircraft Configurat Design Xian 710072 Shaanxi Peoples R China

The precise identification of aerodynamic parameters for aircraft is increasingly critical in aircraft design endeavors, particularly for furnishing accurate dynamic models for ground and flight simulation. To mitigate the dependence on traditional identification methods that rely heavily on aircraft models, this work introduces an aerodynamic parameters identification methodology based on Physics-informed Neural Networks (PINNs). By utilizing the six degrees of freedom motion equation as the physical constraint with the neural network, and the aerodynamic parameters to be identified as the neural network variables, the neural network model is trained to serve as a surrogate for the aircraft model. As a case study, the longitudinal motion of the aircraft is employed to identify and analyze the aerodynamic parameters. A comprehensive comparison is conducted among a method based on the genetic algorithm, a conventional neural network-based approach, and the proposed PINN-based methodology. The results obtained demonstrate that the proposed method can effectively mitigate system and data errors, exhibiting high precision in parameter identification and anti-interference capabilities with regard to noise in data. This innovative approach holds the potential to substantially decrease the reliance on flight test data for parameter identification purposes.

关键词：

来源：评论

学校读者我要写书评

暂无评论

SiamDF: Tracking training data-free siamese tracker

引用

NEURAL NETWORKS 2023年第1期165卷 705-720页

作者： Cai, Huayue Lan, Long Zhang, Jing Zhang, Xiang Luo, Zhigang Natl Univ Def Technol Inst Quantum Changsha Peoples R China Natl Univ Def Technol State Key Lab High Performance Comp Changsha Peoples R China Univ Sydney Sch Comp Sci Sydney Australia Natl Univ Def Technol Lab Digitizing Software Frontier Equipment Changsha Peoples R China Natl Univ Def Technol Coll Comp Sci & Technol Parallel & Distributed Lab Changsha Peoples R China

Much progress has been made in siamese tracking, primarily benefiting from increasing huge training data. However, very little attention has been really paid to the role of huge training data in learning an effective siamese tracker. In this study, we undertake an in-depth analysis of this issue from a novel optimization perspective, and observe that training data is particularly adept at background suppression, thereby refining target representation. Inspired by this insight, we present a data-free siamese tracking algorithm named SiamDF, which requires only a pre-trained backbone and no further fine-tuning on additional training data. Particularly, to suppress background distractors, we separately improve two branches of siamese tracking by retaining the pure target region as target input with the removal of template background, and by exploring an efficient inverse transformation to maintain the constant aspect ratio of target state in search region. Besides, we further promote the center displacement prediction of the entire backbone by eliminating its spatial stride deviations caused by convolution-like quantification operations. Our experimental results on several popular benchmarks demonstrate that SiamDF, free from both offline fine-tuning and online update, achieves impressive performance compared to well-established unsupervised and supervised tracking methods.& COPY;2023 Elsevier Ltd. All rights reserved.

关键词： Siamese tracking Tracking training data-free Pre-training Scale estimation Sharing computation

来源：评论

学校读者我要写书评

暂无评论

Feature Fusion Distillation 1

引用

31st International Conference on Artificial Neural Networks (ICANN)

作者： Tan, Chao Liu, Jie Natl Univ Def Technol Sci & Parallel & Distributed Proc Lab Changsha 410073 Peoples R China Natl Univ Def Technol Lab Software Engn Complex Syst Changsha 410073 Peoples R China

ISBN: (数字)9783031159374

ISBN: (纸本)9783031159374;9783031159367

Most recent efforts made in knowledge distillation (KD) can be credited to filling the representation gap between the cumbersome teacher and its light student. In general, the soft targets, the intermediate feature representation in hidden layers, or a couple of them from the teacher serve as the supervisory signal to educate the student. However, previous works aligned hidden layers one on one and cannot make full use of rich context knowledge. To this end, we propose a Feature Fusion Module (FFM) to concatenate diverse feature maps from different layers to aggregate knowledge as the to-be-distilled dark knowledge. Moreover, to hedge the adverse effects of the fused feature maps, we devise an Asymmetric Switch Function (ASF) to make the transfer process more reliable. The combination of FFM and ASF is termed Feature Fusion Distillation (FFD). Experiments of image classification, object detection, and semantic segmentation on individual benchmarks show FFD jointly assist the student in achieving encouraging performance. It is worth mentioning that when the teacher is ResNet34, the ultimately educated student ResNet18 achieves 71.40% top-1 accuracy on ImageNet-1K.

关键词： Neural network compression Knowledge distillation Knowledge transfer

来源：评论

学校读者我要写书评

暂无评论

Evaluating mesh quality with graph neural networks

引用

ENGINEERING WITH COMPUTERS 2022年第5期38卷 4663-4673页

作者： Wang, Zhichao Chen, Xinhai Li, Tieju Gong, Chunye Pang, Yufei Liu, Jie Natl Univ Def Technol Sci & Technol Parallel & Distributed Proc Lab Changsha 410073 Peoples R China Natl Univ Def Technol Lab Software Engn Complex Syst Changsha 410073 Peoples R China China Aerodynam Res & Dev Ctr Mianyang 621000 Sichuan Peoples R China

The quality of the finite element mesh has a considerable effect on the efficiency and accuracy of computational fluid dynamics (CFD) simulations. To ensure the generated mesh is of good quality, many quality metrics have been proposed to assess the generated mesh, such as aspect ratio, skewness, Jacobian ratio, etc. Such metrics, however, are primarily employed to detect locally distorted mesh elements. There are still no justifiable thresholds for determining whether the generated mesh is of sufficient quality for simulation. Consequently, it is necessary for the professionals to assess the generated mesh afterward which is time-consuming and labor-intensive work. With the ability to learn features on the graph, the graph neural networks have been successfully applied in many application areas to reduce human-computer interaction. In this paper, we define mesh quality evaluation as a graph classification problem. We first propose a novel and sparse-implemented algorithm to transform the mesh data into graph data. We then introduce a deep graph neural network, GMeshNet, to evaluate the mesh quality. Experimental results on the NACA-Market and NACA6510 mesh datasets demonstrate the effectiveness of our proposed network.

关键词： Computational fluid dynamics Mesh quality evaluation Graph neural network Deep learning application

来源：评论

学校读者我要写书评

暂无评论

Three-Dimensional Tradeoffs for Consensus Algorithms: A Review

引用

IEEE TRANSACTIONS ON NETWORK AND SERVICE MANAGEMENT 2022年第2期19卷 1216-1228页

作者： Li, Linhui Shi, Peichang Fu, Xiang Chen, Peng Zhong, Tao Kong, Jinzhu Natl Univ Def Technol Coll Comp Sci Natl Key Lab Parallel & Distributed Proc Changsha 410073 Peoples R China Natl Univ Def Technol Coll Comp Sci Key Lab Software Engn Complex Syst Changsha 410073 Peoples R China Zhejiang Lab Hangzhou 311100 Peoples R China Kylin Software Co Ltd Changsha 410073 Peoples R China

Blockchain has been applied in many fields to solve the problems of trust, security, efficiency benefiting from its tamper-proof and traceability of data. However, it is still necessary to consider the technical constraints that limit the large-scale application of blockchain: scalability, security, and decentralization cannot be achieved altogether. Consensus algorithm is the core of blockchain, which determines the performance of blockchain system to a certain extent. The existing reviews or surveys mainly focus on processes of consensus algorithms, but fall short in covering the current trends and scenarios, thereby lacking intrinsic understanding of their design philosophy. In this paper, we propose a multi-dimensional tradeoff model and unearth various indicators of different dimensions to guide the construction of consensus algorithms. To summarize the existing efforts, we compare and analyze various classical consensus algorithms, and focus on the design principles of these algorithms under the multi-dimensional tradeoff model. According to different requirements, each algorithm has different tradeoffs. Furthermore, we provide different solutions for blockchain in different dimensions. Finally, we summarize the development trend of consensus design and the key technology prospects of blockchain. This is, to the best of our knowledge, the first survey that accomplishes such goals.

关键词： Blockchains Security Consensus algorithm Scalability Throughput Bitcoin Technological innovation Scalability decentralization security consensus

来源：评论

学校读者我要写书评

暂无评论

RF-Siamese: Approaching Accurate RFID Gesture Recognition With One Sample

引用

IEEE TRANSACTIONS ON MOBILE COMPUTING 2024年第1期23卷 797-811页

作者： Ma, Zijing Zhang, Shigeng Liu, Jia Liu, Xuan Wang, Weiping Wang, Jianxin Guo, Song Cent South Univ Sch Comp Sci & Engn Changsha 410017 Peoples R China Cent South Univ Sch Comp Sci & Engn Sci & Technol Parallel & Distributed Proc Lab PDL Changsha 410017 Peoples R China Nanjing Univ State Key Lab Novel Software Technol Nanjing 210093 Peoples R China Hunan Univ Coll Comp Sci & Elect Engn Changsha 410082 Peoples R China Hong Kong Polytech Univ Dept Comp Hong Kong Peoples R China

Performing accurate sensing in diverse environments is a challenging issue in wireless sensing technologies. Existing solutions usually require collecting a large number of samples to train a classifier for every environment, or further assume similar sample distribution between different environments such that a model trained in one environment can be transferred to another. In this paper, we propose RF-Siamese, an RFID-based gesture sensing approach that achieves comparable accuracy to existing solutions but requires only a few samples in each eivironment. RF-Siamese leverages Siamese networks to distinguish different gestures with only a small number of samples and is enhanced by several novel designs to achieve high accuracy in diverse environments. First, the network structure and parameters (e.g., loss function and distance metric) are carefully designed to be suitable for RFID gesture recognition. Second, a permutation-based dataset generation strategy is proposed to make full use of the collected samples to enhance the recognition accuracy. Third, a template matching method is proposed to extend the Siamese network to classify multiple gestures. Extensive experiments on commercial RFID devices demonstrate that RF-Siamese achieves a high accuracy of 0.93 with only one sample of each gesture when recognizing 18 different gestures, while state-of-the-art approaches based on transfer learning and meta learning achieve an accuracy of only 0.59 and 0.70, respectively.

关键词： Sensors Gesture recognition Radiofrequency identification Wireless sensor networks Wireless communication Transfer learning Mobile computing RFID few-shot learning siamese network

来源：评论

学校读者我要写书评

暂无评论

DeepEnhancer: Temporally Consistent Focal Transformer for Comprehensive Video Enhancement 24

DeepEnhancer: Temporally Consistent Focal Transformer for Co...

引用

4th Annual International Conference on Multimedia Retrieval (ICMR)

作者： Jiang, Qin Wang, Qinglin Chi, Lihua Ma, Wentao Li, Feng Liu, Jie Natl Univ Def Technol Lab Digitizing Software Frontier Equipment Sci & Technol Parallel & Distributed Proc Lab Changsha Peoples R China Hunan GuoKe Computil Technol Co Ltd Changsha Peoples R China Anhui Agr Univ Sch Informat & Artificial Intelligence Hefei Peoples R China Hefei Univ Technol Hefei Peoples R China

ISBN: (纸本)9798400706028

Restoring and colorizing old films is a comprehensive video enhancement task, marked by the presence of heterogeneous and structured degradations. Our DeepEnhancer addresses this challenge through a unified pipeline that combines restoration and colorization. In this workflow, we implement a bidirectional propagation strategy. Specifically, we incorporate second-order feature alignment to reduce the accumulation of inaccuracies in optical flow estimation. Simultaneously, we utilize cross-scale long-term attention mechanisms to model correlations within hidden states, thereby ensuring spatial and temporal consistency. To address the notable content loss in aging films, we introduce a temporally consistent focal transformer guided by global information. This transformer utilizes various window levels with distinct sub-window sizes to seamlessly integrate fine-grained and coarse-grained features. Comprehensive experimental results conclusively demonstrate the superiority of our model in both quantitative and qualitative comparisons when compared to existing approaches.

关键词： video restoration colorization Focal Transformer temporal consistency

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：