检索结果-内蒙古大学图书馆

IEEE International Conference on Big Data and Cloud computing (BdCloud)

作者： Chenjing Gao Junli Wang Chungang Yan Key Laboratory of Embedded System and Service Computing (Tongji University) Ministry of Education National (Province-Ministry Joint) Collaborative Innovation Center for Financial Network Security Tongji University Shanghai China

Traffic matrix is the main research object of traffic prediction in software-defined networking. Accurate and timely traffic matrix prediction plays an important role in avoiding network congestion. While various methods have been proposed in previous studies to solve the SDN traffic prediction problem, few of them consider both inter- and intra-flow features. In this paper, we propose a dual-stage attention based Traffic Prediction method for the SDN traffic matrix prediction task. First, our method uses temporal pattern attention as the inter-flow attention mechanism before the encoding stage to realize adaptive feature extraction. Then, temporal attention is introduced as the intra-flow attention mechanism before the decoding stage to capture long-term temporal dependencies. Finally, we use the autoregressive module to handle the highly dynamic SDN traffic volume. Experimental results show that our proposed SDN traffic prediction method can capture more traffic features and improve the SDN traffic matrix prediction performance to some extent.

关键词： Network topology Time series analysis Prediction methods Machine learning Feature extraction Graph neural networks Encoding

来源：评论

学校读者我要写书评

暂无评论

LIMOT: A Tightly-Coupled system for LiDAR-Inertial Odometry and Multi-Object Tracking

arXiv

引用

arXiv 2023年

作者： Zhu, Zhongyang Zhao, Junqiao Huang, Kai Tian, Xuebo Lin, Jiaye Ye, Chen Department of Computer Science and Technology School of Electronics and Information Engineering Tongji University Shanghai China The MOE Key Lab of Embedded System and Service Computing Tongji University Shanghai China Institute of Intelligent Vehicles Tongji University Shanghai China The School of Surveying and Geo-Informatics Tongji University Shanghai China

Simultaneous localization and mapping (SLAM) is critical to the implementation of autonomous driving. Most LiDAR-inertial SLAM algorithms assume a static environment, leading to unreliable localization in dynamic environments. Moreover, the accurate tracking of moving objects is of great significance for the control and planning of autonomous vehicles. This study proposes LIMOT, a tightly-coupled multi-object tracking and LiDAR-inertial odometry system that is capable of accurately estimating the poses of both ego-vehicle and objects. We propose a trajectory-based dynamic feature filtering method, which filters out features belonging to moving objects by leveraging tracking results before scan-matching. Factor graph-based optimization is then conducted to optimize the bias of the IMU and the poses of both the ego-vehicle and surrounding objects in a sliding window. Experiments conducted on the KITTI tracking dataset and self-collected dataset show that our method achieves better pose and tracking accuracy than our previous work DL-SLOT and other baseline methods. Our open-source implementation is available at https://***/tiev-tongji/LIMOT. Copyright © 2023, The Authors. All rights reserved.

关键词： Autonomous vehicles

来源：评论

学校读者我要写书评

暂无评论

Learning Sequence Descriptor based on Spatio-Temporal Attention for Visual Place Recognition

arXiv

引用

arXiv 2023年

作者： Zhao, Junqiao Zhang, Fenglin Cai, Yingfeng Tian, Gengxuan Mu, Wenjie Ye, Chen Feng, Tiantian Department of Computer Science and Technology School of Electronics and Information Engineering Tongji University Shanghai China The MOE Key Lab of Embedded System and Service Computing Tongji University Shanghai China Institute of Intelligent Vehicles Tongji University Shanghai China School of Surveying and Geo-Informatics Tongji University Shanghai China

Visual Place Recognition (VPR) aims to retrieve frames from a geotagged database that are located at the same place as the query frame. To improve the robustness of VPR in perceptually aliasing scenarios, sequence-based VPR methods are proposed. These methods are either based on matching between frame sequences or extracting sequence descriptors for direct retrieval. However, the former is usually based on the assumption of constant velocity, which is difficult to hold in practice, and is computationally expensive and subject to sequence length. Although the latter overcomes these problems, existing sequence descriptors are constructed by aggregating features of multiple frames only, without interaction on temporal information, and thus cannot obtain descriptors with spatio-temporal discrimination. In this paper, we propose a sequence descriptor that effectively incorporates spatio-temporal information. Specifically, spatial attention within the same frame is utilized to learn spatial feature patterns, while attention in corresponding local regions of different frames is utilized to learn the persistence or change of features over time. We use a sliding window to control the temporal range of attention and use relative positional encoding to construct sequential relationships between different features. This allows our descriptors to capture the intrinsic dynamics in a sequence of frames. Comprehensive experiments on challenging benchmark datasets show that the proposed approach outperforms recent state-of-the-art methods. The code is available at https://***/tiev-tongji/Spatio-Temporal-SeqVPR. Copyright © 2023, The Authors. All rights reserved.

关键词： Query processing

来源：评论

学校读者我要写书评

暂无评论

Memory-Based Contrastive Learning with Optimized Sampling for Incremental Few-Shot Semantic Segmentation

Memory-Based Contrastive Learning with Optimized Sampling fo...

引用

IEEE International Symposium on Circuits and systems (ISCAS)

作者： Yuxuan Zhang Miaojing Shi Taiyi Su Hanli Wang Department of Computer Science & Technology Tongji University Shanghai P. R. China Key Laboratory of Embedded System & Service Computing Ministry of Education Tongji University Shanghai P. R. China Department of Control Science & Engineering Tongji University Shanghai P. R. China

ISBN: (数字)9798350330991

ISBN: (纸本)9798350331004

Incremental few-shot semantic segmentation (IFSS) aims to incrementally expand a semantic segmentation model’s ability to identify new classes based on few samples. However, it grapples with the dual challenges of catastrophic forgetting (due to feature drift in old classes) and overfitting (triggered by inadequate samples in new classes). To address these issues, a novel approach is proposed to integrate pixel-wise and region-wise contrastive learning, complemented by an optimized example and anchor sampling strategy. The proposed method incorporates a region memory and pixel memory designed to explore the high-dimensional embedding space more effectively. The memory, retaining the feature embeddings of known classes, facilitates the calibration and alignment of seen class features during the learning process of new classes. To further mitigate overfitting, the proposed approach implements an optimized example and anchor sampling strategy. Extensive experiments show the competitive performance of the proposed method. The source code of this work can be found in https://***.

关键词： Circuits and systems Semantic segmentation Source coding Semantics Calibration

来源：评论

学校读者我要写书评

暂无评论

LAMPT: LAbel Mask-Predicted Transformer for Extreme Multi-label Text Classification

LAMPT: LAbel Mask-Predicted Transformer for Extreme Multi-la...

引用

IEEE International Conference on Big Data

作者： Yifan Xu Minghao Zhu Junli Wang Chungang Yan Key Laboratory of Embedded System and Service Computing (Tongji University) Ministry of Education. National (Province-Ministry Joint) Collaborative Innovation Center for Financial Network Security Tongji University Shanghai China

ISBN: (纸本)9781665480468

Extreme Multi-label text Classification ( XMC) is a task of recalling the most relevant labels for each given text from an extremely large-scale label set. It is emphasized that XMC is a more complex classification task because there are two main problems: large space of labels and the labels in XMC tasks tend to be correlated. Existing methods attempt to model label correlations by viewing the XMC task as a sequence generation problem however they still suffer from (1) using the slow serial decoding strategy where labels are predicted one-by-one. (2) needing to compare a mass of label ordering strategies in the decoding stage to achieve satisfied accuracy. In this work, we propose LAbel Mask-Predicted Transformer (LAMPT) to address the both issues, which is a novel non-autoregressive generation model that (1) enriches the input raw text representation with the additional label features by fully exploiting the label dependencies, (2) allows for efficient parallel decoding thanks to its non-autoregressive decoding formulation and mask-prediced training strategy. Experimental results demonstrate that single model performance is substantially enhanced by LAMPT. On a Wiki dataset with thirty-one thousand labels, LAMPT-XLNet accuracy has gained 1.6% relative improvement on P@3 over the LightXML-XLNet. Also, the P@1 of ensemble LAMPT is 90.00%, a significant i mprovement over the state-of-the-art ensemble LightXML (transformer-based) and AttentionXML (LSTM-based), which achieve 89.45% and 87.47%, respectively.

关键词： Training Correlation Text categorization Information services Big Data Predictive models Transformers

来源：评论

学校读者我要写书评

暂无评论

Cross-Modal Representation Reconstruction for Zero-Shot Classification

Cross-Modal Representation Reconstruction for Zero-Shot Clas...

引用

IEEE International Conference on Acoustics, Speech and Signal Processing

作者： Yu Wang Shenjie Zhao Key Laboratory of Embedded System and Service Computing Ministry of Education P. R. China Mojie Technology (Ningbo)

ISBN: (纸本)9781728176055;9781728176062

Zero-shot learning (ZSL) aims to recognize novel classes without training samples through transferring knowledge from seen classes, based on the assumption that both the seen and unseen classes share a latent semantic space. Previous works either focus on directly learning various mapping functions between visual space and semantic space, or searching a latent common subspace to alleviate semantic gap between different modalities. However, few methods directly learn modality-invariant representations for ZSL. In this paper, we propose a Cross-Modal Representation Reconstruction (CM-RR) framework to bridge the semantic gap between visual features and semantic attributes, as well as introducing a novel regularizer for automatically feature selection. Moreover, an iterative optimization process based on the ALM (Augmented Lagrangian Method) algorithm with the alternating direction strategy is developed to solve the proposed formulation. Extensive experiments on four benchmark datasets show the effectiveness of the proposed approach, and even the performance surpasses some deep learning based methods.

关键词： Training Visualization Semantics Signal processing algorithms Signal processing Feature extraction Iterative algorithms

来源：评论

学校读者我要写书评

暂无评论

Cross-Domain Semantic Segmentation of Urban Scenes via Multi-Level Feature Alignment

Cross-Domain Semantic Segmentation of Urban Scenes via Multi...

引用

International Conference on Pattern Recognition

作者： Bin Zhang Shengjie Zhao Rongqing Zhang Key Laboratory of Embedded System and Service Computing Ministry of Education China School of Software Engineering Tongji University Shanghai China

Semantic segmentation is an essential task in plenty of real-life applications such as virtual reality, video analysis, autonomous driving, etc. Recent advancements in fundamental vision-based tasks ranging from image classification to semantic segmentation have demonstrated deep learning-based models' high capability in learning complicated representation on large datasets. Nevertheless, manually labeling semantic segmentation dataset with pixel-level annotation is extremely labor-intensive. To address this problem, we propose a novel multi-level feature alignment framework for cross-domain semantic segmentation of urban scenes by exploiting generative adversarial networks. In the proposed multi-level feature alignment method, we first translate images from one domain to another one. Then the discriminative feature representations extracted by the deep neural network are concatenated, followed by domain adversarial learning to make the intermediate feature distribution of the target domain images close to those in the source domain. With these domain adaptation techniques, models trained with images in the source domain where the labels are easy to acquire can be deployed to the target domain where the labels are scarce. Experimental evaluations on various mainstream benchmarks confirm the effectiveness as well as robustness of our approach.

关键词： Image segmentation Solid modeling Semantics Neural networks Virtual reality Feature extraction Robustness

来源：评论

学校读者我要写书评

暂无评论

Design and Implementation of DSP Cache 21

Design and Implementation of DSP Cache

引用

21st IEEE International Conference on Communication Technology, ICCT 2021

作者： Liang, Kunkun Wu, Jun Ren, Haoqi Zhang, Zhifeng Tan, Bin Tongji University Electronics and Information Engineering College Key Laboratory of Embedded System and Service Computing Ministry of Education Shanghai201804 China School of Computer Science Fudan University Shanghai200438 China College of Electronics and Information Engineering Jinggangshan University Ji'an343009 China

ISBN: (纸本)9781665432061

The cache can improve the DSP processor's access speed to external memory and solve the 'Storage Wall' problem. Designing an efficient and flexible cache module plays an vital role in improving the memory access efficiency and overall performance of the DSP. Based on the analysis of DSP storage level design requirements, combined with the actual pipeline structure of the DSP processor SWIFT independently developed by our laboratory, in this paper, we design and implement first- level instruction and data cache. For L1 D-Cache, the relevant conflict detection and processing module is designed to support four parallel Load/Store access requests. The size of the cache can be flexibly configured according to actual application requirements to achieve a balance between power consumption and latency. To write back the modified data left in the cache before the cache size is changed, the Cacheclean instruction is designed. Finally, module-level functional verification and logic synthesis are carried out. The results show that the cache function meets expectations, and the critical path after logic synthesis optimization meets the requirement of 1GHz frenuency. © 2021 IEEE.

关键词： Pipelines

来源：评论

学校读者我要写书评

暂无评论

HeavyTracker: An Efficient Algorithm for Heavy-Hitter Detection in High-Speed Networks

HeavyTracker: An Efficient Algorithm for Heavy-Hitter Detect...

引用

International Conference on Parallel and Distributed systems (ICPADS)

作者： Jiakun Yu Yu-E Sun He Huang Yang Du Guoju Gao Hongli Xu Shigang Chen School of Computer Science and Technology Soochow University Suzhou China School of Rail Transportation Soochow University Suzhou China Ministry of Education Key Laboratory of Embedded System and Service Computing (Tongji University) Shanghai China School of Computer Science and Technology University of Science and Technology of China Hefei China

Detecting heavy hitters that constitute the majority of network traffic is a critical task in network measurement. However, the highly skewed network traffic and the size-limited on-chip memory pose great challenges for heavy-hitter detection. The prior arts either use sketches to track all flows or maintain a fix-sized tracking list for recording elephant flows, resulting in limited detection precision or time-consuming tracking list exchanges. This paper presents HeavyTracker, an efficient algorithm that detects heavy hitters based on a count-with-threshold strategy. A two-dimensional tracker unit array is employed in our design to capture the heavy hitters, where each flow is randomly mapped to one unit in all rows. The tracker unit is designed to accept many flows and precisely report the largest two with frequencies that reach a predefined threshold. This design eliminates the frequent and time-consuming exchanges for maintaining the tracking list, allowing us to use a hash table to track the reported elephant flows. Experimental results based on real Internet traces show that HeavyTracker achieves a high $\mathbf{F}_{\beta}$-Score of threshold-t detection (0.97) and over 99% precision of top-k detection under a tight memory size. Besides, it reduces the frequency estimation error by 97.4% compared to the state-of-the-art.

关键词： Art High-speed networks Measurement uncertainty Telecommunication traffic Size measurement Frequency estimation Internet

来源：评论

学校读者我要写书评

暂无评论

Towards Energy-Efficiency: Integrating MATD3 Reinforcement Learning Method for Computational Offloading in RIS-Aided UAV-MEC Environments

引用

IEEE Internet of Things Journal 2025年

作者： Wu, Liangshun Zhang, Cong Zhang, Bin Du, Jianbo Qu, Junsuo Key Laboratory of Embedded System and Service Computing Tongji University Ministry of Education Shanghai201804 China Shanghai Key Laboratory of Trustworthy Computing East China Normal University Shanghai201210 China School of Electronic Information and Electrical Engineering Shanghai Jiao Tong University Shanghai200240 China Xi'an Key Laboratory of Advanced Control and Intelligent Process School of Automation Xi'an University of Posts & Telecommunications Xi'an710061 China Xinjiang University of Political Science and Law Tumushuke832003 China School of Remote Sensing and Information Engineering Wuhan University Wuhan430079 China Department of Computer City University of Hong Kong 999077 Hong Kong School of Communications and Information Engineering Xi'an University of Posts & Telecommunications Xi'an710061 China

With the proliferation of IoT devices, there is an escalating demand for enhanced computing and communication capabilities. Mobile Edge computing (MEC) addresses this need by relocating computing resources to the network edge, thereby delivering swifter and more efficient services. This paper introduces a computation offloading and energy consumption optimization framework that leverages Reconfigurable Intelligent Surfaces (RIS), Unmanned Aerial Vehicles (UAVs), and MEC. The scheme aims to maximize energy efficiency through the optimization of task allocation, RIS phase shifts, and UAV trajectories. By employing the Multi-Agent Twin Delayed Deep Deterministic Policy Gradient (MATD3) reinforcement learning algorithm, the paper further refines UAV trajectories and RIS configurations. The simulation results indicate that the proposed method surpasses the traditional Concave-Convex Procedure (CCCP) algorithm in both UAV trajectory control and RIS configuration, demonstrating quicker convergence and enhanced stability. The method proves to be adaptable to diverse environments and tasks, showcasing notable benefits in RIS-assisted interference suppression, particularly with large RIS, thereby enhancing UAV data reception rates. Additionally, MATD3 exhibits faster and smoother convergence for extended task durations and smaller RIS scenarios. Simulation results reveal that UAVs tend to move closer to RIS, with energy efficiency falling as IoT tasks increase, affirming the proposed algorithm's high energy efficiency and effectiveness. © 2025 IEEE.

关键词： Unmanned aerial vehicles (UAV)

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：