检索结果-内蒙古大学图书馆

TSCompiler: efficient compilation framework for dynamic-shape models

science China(Information sciences) 2024年第10期67卷 67-84页

作者： Xiang LUO Chen ZHANG Chenbo GENG Yanzhi YI Jiahui HU Renwei ZHANG Zhen ZHANG Gianpietro CONSOLARO Fan YANG Tun LU Ning GU Li SHANG School of Computer Science Fudan University School of Electronic Information and Electrical Engineering Shanghai Jiao Tong University School of Computer Science and Technology Harbin Institute of Technology Huawei Technologies Co. Ltd. Huawei Paris Research Center School of Microelectronics Fudan University

Today's deep learning models face an increasing demand to handle dynamic shape tensors and computation whose shape information remains unknown at compile time and varies in a nearly infinite range at runtime. This shape dynamism brings tremendous challenges for existing compilation pipelines designed for static models which optimize tensor programs relying on exact shape values. This paper presents TSCompiler, an end-to-end compilation framework for dynamic shape models. TSCompiler first proposes a symbolic shape propagation algorithm to recover symbolic shape information at compile time to enable subsequent optimizations. TSCompiler then partitions the shape-annotated computation graph into multiple subgraphs and fine-tunes the backbone operators from the subgraph within a hardware-aligned search space to find a collection of high-performance schedules. TSCompiler can propagate the explored backbone schedule to other fusion groups within the same subgraph to generate a set of parameterized tensor programs for fused cases based on dependence analysis. At runtime, TSCompiler utilizes an occupancy-targeted cost model to select from pre-compiled tensor programs for varied tensor shapes. Extensive evaluations show that TSCompiler can achieve state-of-the-art speedups for dynamic shape models. For example, we can improve kernel efficiency by up to 3.97× on NVIDIA RTX3090, and 10.30× on NVIDIA A100 and achieve up to five orders of magnitude speedups on end-to-end latency.

关键词： machine learning tensor compilers dynamic shape operator fusion code generation auto-tuning

来源：评论

学校读者我要写书评

暂无评论

Domain generalization with semi-supervised learning for people-centric activity recognition

引用

science China(Information sciences) 2025年第1期68卷 171-188页

作者： Jing LIU Wei ZHU Di LI Xing HU Liang SONG Academy for Engineering & Technology Fudan University Shanghai East-bund Research Institute on Networking Systems of AI School of Optoelectronic Information and Computer Engineering University of Shanghai for Science & Technology

People-centric activity recognition is one of the most critical technologies in a wide range of real-world applications,including intelligent transportation systems, healthcare services, and brain-computer interfaces. Large-scale data collection and annotation make the application of machine learning algorithms prohibitively expensive when adapting to new tasks. One way of circumventing this limitation is to train the model in a semi-supervised learning manner that utilizes a percentage of unlabeled data to reduce the labeling burden in prediction tasks. Despite their appeal, these models often assume that labeled and unlabeled data come from similar distributions, which leads to the domain shift problem caused by the presence of distribution gaps. To address these limitations, we propose herein a novel method for people-centric activity recognition,called domain generalization with semi-supervised learning(DGSSL), that effectively enhances the representation learning and domain alignment capabilities of a model. We first design a new autoregressive discriminator for adversarial training between unlabeled and labeled source domains, extracting domain-specific features to reduce the distribution gaps. Second, we introduce two reconstruction tasks to capture the task-specific features to avoid losing information related to representation learning while maintaining task-specific consistency. Finally, benefiting from the collaborative optimization of these two tasks, the model can accurately predict both the domain and category labels of the source domains for the classification task. We conduct extensive experiments on three real-world sensing datasets. The experimental results show that DGSSL surpasses the three state-of-the-art methods with better performance and generalization.

关键词： activity recognition deep learning domain generalization semi-supervised learning adversarial training

来源：评论

学校读者我要写书评

暂无评论

Multi-Dimensional Training Optimization for Efficient Federated Synergy Learning

引用

IEEE Transactions on Mobile Computing 2025年第7期24卷 6243-6258页

作者： Fu, Shucun Dong, Fang Chen, Runze Shen, Dian Zhang, Jinghui He, Qiang Southeast University School of Computer Science and Engineering Nanjing China Huazhong University of Science and Technology National Engineering Research Center for Big Data Technology and System Services Computing Technology and System Lab Cluster and Grid Computing Lab School of Computer Science and Technology Wuhan China

Edge learning (EL) is an end-to-edge collaborative learning paradigm enabling devices to participate in model training and data analysis, opening countless opportunities for edge intelligence. As a promising EL framework, federated synergy learning (FSyL) mitigates the computation and communication overhead on resource-constrained devices by offloading partial model layers to the edge server for synergistic training. Nevertheless, due to the system and statistical heterogeneity, naively using existing FSyL methods is significantly time-consuming and causes accuracy degradation. Motivated by this issue, this paper introduces a novel FSyL framework that integrates multi-dimensional training optimization and formulates the edge learning cost minimization (ELCM) problem. To tackle the ELCM efficiently, we design OL-MG, an OnLine Model Splitting and Resource Provisioning Game. Specifically, we first reformulate and decompose the original ELCM based on data quality evaluation. Then, given a model splitting decision, we determine the optimal resource provisioning in Sub-problem1, based on which optimal model splitting in Sub-problem2 is modeled as a potential game. Subsequently, we introduce a decentralized algorithm to find a Nash equilibrium (NE) solution. Furthermore, we further extend OL-MG to support a budget-aware multi-edge scenario. Extensive experiments demonstrate that the proposed mechanism significantly outperforms state-of-the-art methods in cost-saving and accuracy improvement. © 2025 IEEE.

关键词： Federated learning

来源：评论

学校读者我要写书评

暂无评论

RoleNet: A multiple features fusion network for role classification in cantonese opera

引用

Multimedia Tools and Applications 2025年 1-14页

作者： Li, Yue Peng, Zhengwei Xu, Di Chen, Yuanguang Chen, Guoan School of Computer Science and Engineering South China University of Technology Guangdong Guangzhou510006 China School of Computer Science and Engineering Sun Yat-sen University Guangdong Guangzhou510006 China

Cantonese opera, a key facet of Chinese traditional opera, boasts profound cultural and artistic value and has been designated as intangible cultural heritage. The use of certain roles is a basic concept in Cantonese opera, where each role has a specific style of singing, movement, and costume that performers are trained to perform throughout their careers. Therefore, identifying the role category of characters in a play can provide theoretical and systematic foundations for further researches and artistic explorations. By dissecting musical traits and performance styles of each role, comprehensive studies on its regional and artistic nuances are enabled. To achieve role classification in Cantonese opera, we propose RoleNet, an integration network that consists of SincNets, transformers, and a feature fusion block. For a given musical fragment (e.g., an audio signal), SincNets extract 1D features at multiple scales and transformers extract features from time and frequency axes from the 2D Mel Spectrogram. Subsequently, the extracted features are concatenated by the fusioner using multiple features selection strategy to perform role classification tasks. The experimental results on a real-world dataset demonstrated the superior performance of RoleNet compared to single-objective methods. An overall classification accuracy of 98.67% was achieved. The codes are available at https://***/AaronPeng920/RoleNet. © The Author(s), under exclusive licence to Springer science+Business Media, LLC, part of Springer Nature 2025.

关键词： Feature Selection

来源：评论

学校读者我要写书评

暂无评论

A review on production and application of direct reduced iron in gas-based shaft furnace–electric arc furnace route

引用

Journal of Iron and Steel Research International 2025年第3期32卷 485-518页

作者： Ling-zhi Yang Zeng Feng Hang Hu Guang-sheng Wei Bo-tao Xue Yu-feng Guo Tao Jiang School of Minerals Processing and Bioengineering Central South UniversityChangsha410083HunanChina School of Metallurgical and Ecological Engineering University of Science and Technology BeijingBeijing100083China

The iron and steel industry,standing as a quintessential manufacture example with high consumption,pollution and emissions,faces significant environmental and sustainable development *** arc furnace(EAF)steelmaking process mainly uses scrap as raw material and is characterized by environmentally friendly and recyclable ***,the further development of EAF route in China is limited by the reserve,supply,availability and quality of scrap *** reduced iron(DRI)is one of typical low-carbon and clean charges,which can effectively make up for the adverse effects caused by the lack of *** physical and chemical properties,classifications,and production technologies of DRI are firstly *** particular,the reducing gas types,reduction temperature,and reduction mechanism of the DRI production with gas-based shaft furnace(SF)technology are *** the crucial role played by DRI application in EAF,the influences of DRI addition on EAF smelting rules and operations including the blending and charging process,heat transfer and melting in molten bath,slag formation operation,refractory corrosion,and slag system evolution are then further ***,the comparative analysis and assessment of the consumption level of material and energy as well as the cleaner production both covering the clean chemical composition of molten steel and the clean environment impact in EAF steelmaking with DRI charged are *** perspectives of metallurgical process engineering,a suitable route of hydrogen generation and application(from coke oven gas,methanol,and clean energy power),CO_(2) capture and utilization integrated with SF–EAF process is *** view of the difficulties in large-scale DRI application in EAF,the follow-up work should focus on the investigation of DRI charging and melting,slag system evolution and molten pool reaction rules,as well as the developments of the DRI standardized use technology and intell

关键词： Electric arc furnace steelmaking Direct reduced iron Hydrogen-based shaft furnace Sustainable production Low-carbon development

来源：评论

学校读者我要写书评

暂无评论

Exploring Unbiased Activation Maps for Weakly Supervised Tissue Segmentation of Histopathological Images

引用

IEEE Transactions on Medical Imaging 2025年第6期44卷 2631-2642页

作者： Kang, Yuxin Li, Hansheng Shi, Xiaoshuang Zhang, Xiao Xing, Yaqiong Wen, Yuting Wang, Yi Cui, Lei Feng, Jun Yang, Lin Yulin University Department of Information Engineering Shaanxi Yulin China The School of Information Science and Technology Northwest University Shaanxi Xi'an China Northwest University School of Information Science and Technology Shaanxi Xi'an China University of Electronic Science and Technology of China Department of Computer Science and Engineering Sichuan Chengdu China Xi'An Jiaotong University School of Computer Science and Technology Shaanxi Xi'an China

Tissue segmentation in histopathological images plays a crucial role in computational pathology, owing to its significant potential to indicate the prognosis of cancer patients. Presently, numerous Weakly Supervised Semantic Segmentation (WSSS) methods strive to utilize image-level labels to achieve pixel-level segmentation, aiming to minimize the need for detailed annotations. Most of these methods rely on Class Activation Maps (CAM) extracted from classification models, frequently leading to poor coverage of objects. The major cause is attributed to the strong inductive bias of the classification model, focusing primarily on discriminative feature of objects, rather than non-discriminative features. Inspired by this, we propose a simple yet effective method that introduces a self-supervised task by exploiting both the discriminative and non-discriminative features, and generate Unbiased Activation Maps (UAM) to encompass the whole object. Specifically, our method entails clustering all spatial features of an object class to derive semantic centers. Each center then works as a spatial filter that amplifies similar feature and suppresses dissimilar feature, and extract high-quality pseudo-labels (some noise at object boundaries). Moreover, we further propose a Noise-Reduced (NR) Learning method to train the segmentation network towards credible signals and lessen the impact of false predictions. Comprehensive experimental results on two public histopathology image datasets demonstrate the superior performance of our method over the state-of-the-art weakly supervised segmentation methods. © 2025 IEEE. All rights reserved.

关键词： Semantic Segmentation

来源：评论

学校读者我要写书评

暂无评论

Real-time holographic camera for obtaining real 3D scene hologram

引用

Light(science & Applications) 2025年第3期14卷 789-799页

作者： Zhao-Song Li Chao Liu Xiao-Wei Li Yi Zheng Qian Huang Yi-Wei Zheng Ye-Hao Hou Chen-Liang Chang Da-Wei Zhang Song-Lin Zhuang Di Wang Qiong-Hua Wang School of Instrumentation and Optoelectronic Engineering Beihang UniversityBeijing100191China School of Optical-Electrical and Computer Engineering University of Shanghai for Science and TechnologyShanghai 200093China

As a frontier technology,holography has important research values in fields such as bio-micrographic imaging,light feld modulation and data ***,the real-time acquisition of 3D scenes and high-fidelity reconstruction technology has not yet made a breakthrough,which has seriously hindered the development of ***,a novel holographic camera is proposed to solve the above inherent problems *** proposed holographic camera consists of the acquisition end and the calculation *** the acquisition end of the holographic camera,specially configured liquid materials and liquid lens structure based on voice-coil motor-driving are used to produce the liquid camera,so that the liquid camera can quickly capture the focus stack of the real 3D scene within 15 *** the calculation end,a new structured focus stack network(FS-Net)is designed for hologram *** training the FS-Net with the focus stack renderer and learnable Zernike phase,it enables hologram calculation within 13 *** the first device to achieve real-time incoherent acquisition and high-fidelity holographic reconstruction of a real 3D scene,our proposed holographic camera breaks technical bottlenecks of difficulty in acquiring the real 3D scene,low quality of the holographic reconstructed image,and incorrect defocus *** experimental results demonstrate the effectiveness of our holographic camera in the acquisition of focal plane information and hologram calculation of the real 3D *** proposed holographic camera opens up a new way for the application of holography in fields such as 3D display,light field modulation,and 3D measurement.

关键词： holographic camera feld modulation hologram calculation liquid lens structured focus stack network D scene reconstruction d scenes real time acquisition

来源：评论

学校读者我要写书评

暂无评论

High Dynamic Range Imaging for Dynamic Scenes Based on Multi-Level Spike Camera

引用

IEEE Transactions on Circuits and Systems for Video technology 2025年第6期35卷 5394-5406页

作者： Zhu, Zhenkun Xiong, Ruiqin Zhao, Jing Zhao, Rui Fan, Xiaopeng Zhu, Shuyuan Huang, Tiejun Peking University School of Computer Science Beijing100871 China Harbin Institute of Technology Department of Computer Science and Technology Harbin150001 China University of Electronic Science and Technology of China School of Information and Communication Engineering Chengdu611731 China

Spike camera is a retina-inspired neuromorphic camera which can capture dynamic scenes of high-speed motion by firing a continuous stream of spikes at an extremely high temporal resolution. The limitation in the current design is that each spike only represents the arrival of a fixed amount of photons. It can not deal with strong light areas in which the amount of accumulated photons reaches the pre-specified threshold multiple times within a single readout interval. In this paper, we propose a new spike camera model of high-speed imaging for high dynamic range scenarios. In this scheme, each pixel accumulates the incoming photons persistently and generates a new type of spike stream in which each spike symbol may be associated with different levels, indicating the arrival of different amounts of photons since the last readout. This enables the camera to support dynamic scenes with wider dynamic range. To achieve this, we propose a two-level buffer mechanism, one for photon accumulation and one for spike-firing encoding. We use a register to hold the number of spike-firings which has not been read out yet. At each readout time, the major part in the counter is read out via a carefully designed exponential encoding and the counter is updated. Such encoding and readout strategy enables a very efficient expansion of the dynamic range using a small number of encoding bits. Furthermore, we propose an image reconstruction scheme for the proposed camera, utilizing both spike intervals and spike levels to recover the light intensity. We incorporate Mamba and propose a temporal-spatial selective scan mechanism to extract temporal-spatial correlation within spike streams. We employ a pyramid adaptive filtering and alignment module to achieve coarse-to-fine feature alignment. Experimental results show that the proposed scheme can achieve better imaging quality and outperform the existing spike camera in high dynamic range scenarios. © 1991-2012 IEEE.

关键词： Adaptive filtering

来源：评论

学校读者我要写书评

暂无评论

YOLO-LFD: A Lightweight and Fast Model for Forest Fire Detection

引用

computers, Materials & Continua 2025年第2期82卷 3399-3417页

作者： Honglin Wang Yangyang Zhang Cheng Zhu School of Artificial Intelligence Nanjing University of Information Science and TechnologyNanjing210044China School of Computer Science Nanjing University of Information Science and TechnologyNanjing210044China Electrical&Computer Engineering University of Illinois at Urbana-ChampaignUrbanaIL 61801USA

Forest fires pose a serious threat to ecological balance, air quality, and the safety of both humans and wildlife. This paper presents an improved model based on You Only Look Once version 5 (YOLOv5), named YOLO Lightweight Fire Detector (YOLO-LFD), to address the limitations of traditional sensor-based fire detection methods in terms of real-time performance and accuracy. The proposed model is designed to enhance inference speed while maintaining high detection accuracy on resource-constrained devices such as drones and embedded systems. Firstly, we introduce Depthwise Separable Convolutions (DSConv) to reduce the complexity of the feature extraction network. Secondly, we design and implement the Lightweight Faster Implementation of Cross Stage Partial (CSP) Bottleneck with 2 Convolutions (C2f-Light) and the CSP Structure with 3 Compact Inverted Blocks (C3CIB) modules to replace the traditional C3 modules. This optimization enhances deep feature extraction and semantic information processing, thereby significantly increasing inference speed. To enhance the detection capability for small fires, the model employs a Normalized Wasserstein Distance (NWD) loss function, which effectively reduces the missed detection rate and improves the accuracy of detecting small fire sources. Experimental results demonstrate that compared to the baseline YOLOv5s model, the YOLO-LFD model not only increases inference speed by 19.3% but also significantly improves the detection accuracy for small fire targets, with only a 1.6% reduction in overall mean average precision (mAP)@0.5. Through these innovative improvements to YOLOv5s, the YOLO-LFD model achieves a balance between speed and accuracy, making it particularly suitable for real-time detection tasks on mobile and embedded devices.

关键词： Forest fire detection YOLOv5 lightweight small object detection

来源：评论

学校读者我要写书评

暂无评论

Learning from Heterogeneity: A Dynamic Learning Framework for Hypergraphs

IEEE Transactions on Artificial Intelligence

引用

IEEE Transactions on Artificial Intelligence 2025年第6期6卷 1513-1528页

作者： Zhang, Tiehua Liu, Yuze Shen, Zhishu Ma, Xingjun Qi, Peng Ding, Zhijun Jin, Jiong Tongji University School of Computer Science and Technology Shanghai201804 China Ant Group Shanghai200122 China Wuhan University of Technology School of Computer Science and Artificial Intelligence Wuhan430070 China Fudan University School of Computer Science Shanghai200438 China Tongji University Department of Control Science and Engineering Shanghai201804 China Swinburne University of Technology School of Science Computing and Engineering Technologies Melbourne3122 Australia

Graph neural network (GNN) has gained increasing popularity in recent years owing to its capability and flexibility in modeling complex graph structure data. Among all graph learning methods, hypergraph learning is a technique for exploring the implicit higher-order correlations when training the embedding space of the graph. In this article, we propose a hypergraph learning framework named learning from heterogeneity (LFH) that is capable of dynamic hyperedge construction and attentive embedding update utilizing the heterogeneity attributes of the graph. Specifically, in our framework, the high-quality features are first generated by the pairwise fusion strategy that utilizes explicit graph structure information when generating initial node embedding. Afterward, a hypergraph is constructed through the dynamic grouping of implicit hyperedges, followed by the type-specific hypergraph learning process. To evaluate the effectiveness of our proposed framework, we conduct comprehensive experiments on several popular datasets with twelve state-of-the-art models on both node classification and link prediction tasks, which fall into categories of homogeneous pairwise graph learning, heterogeneous pairwise graph learning, and hypergraph learning. The experimental results demonstrate a significant performance gain (an average of 12.9% in node classification and 12.8% in link prediction) compared with recent state-of-the-art methods. © 2020 IEEE.

关键词： Contrastive Learning

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：