检索结果-南通市图书馆

Heterogeneous decentralised machine unlearning with seed model distillation

在线全文

学校读者我要写书评

暂无评论

CAAI Transactions on Intelligence Technology 2024年第3期9卷 608-619页

作者： Guanhua Ye Tong Chen Quoc Viet Hung Nguyen Hongzhi Yin school of Information Technology and Electrical Engineering The University of QueenslandSt LuciaQueenslandAustralia school of Information and communication Technology Griffith UniversityBrisbaneQueenslandAustralia

As some recent information security legislation endowed users with unconditional rights to be forgotten by any trained machine learning model,personalised IoT service pro-viders have to put unlearning functionality into their *** most straight-forward method to unlearn users'contribution is to retrain the model from the initial state,which is not realistic in high throughput applications with frequent unlearning *** some machine unlearning frameworks have been proposed to speed up the retraining process,they fail to match decentralised learning scenarios.A decentralised unlearning framework called heterogeneous decentralised unlearning framework with seed(HDUS)is designed,which uses distilled seed models to construct erasable en-sembles for all ***,the framework is compatible with heterogeneous on-device models,representing stronger scalability in real-world *** experiments on three real-world datasets show that our HDUS achieves state-of-the-art performance.

关键词： data mining data privacy machine learning

Pupillometry Analysis of Rapid Serial Visual Presentation at Five Presentation Rates

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

Tsinghua Science and Technology 2024年第2期29卷 543-552页

作者： Xi Luo Yanfei Lin Rongxiao Guo Xirui Zhao Shangen Zhang Xiaorong Gao school of Integrated Circuits and Electronics Beijing Institute of TechnologyBeijing 100081China school of Computer and communication Engineering University of Science and Technology BeijingBeijing 100083China school of Medicine Tsinghua UniversityBeijing 100084China

In this study,the effect of presentation rates on pupil dilation is investigated for target recognition in the Rapid Serial Visual Presentation(RSVP)*** this experiment,the RSVP paradigm with five different presentation rates,including 50,80,100,150,and 200 ms,is *** pupillometry data of 15 subjects are collected and *** pupillometry results reveal that the peak and average amplitudes for pupil size and velocity at the 80-ms presentation rate are considerably higher than those at other presentation *** average amplitude of pupil acceleration at the 80-ms presentation rate is significantly higher than those at the other presentation *** latencies under 50-and 80-ms presentation rates are considerably lower than those of 100-,150-,and 200-ms presentation ***,no considerable differences are observed in the peak,average amplitude,and latency of pupil size,pupil velocity,and acceleration under 100-,150-,and 200-ms presentation *** results reveal that with the increase in the presentation rate,pupil dilation first increases,then decreases,and later reaches *** 80-ms presentation rate results in the largest point of pupil *** correlation is observed between pupil dilation and recognition accuracy under the five presentation rates.

关键词： pupil dilation presentation rate Rapid Serial Visual Presentation(RSVP)

Precise region semantics-assisted GAN for pose-guided person image generation

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

CAAI Transactions on Intelligence Technology 2024年第3期9卷 665-678页

作者： Ji Liu Zhenyu Weng Yuesheng Zhu communication and Information Security Lab Shenzhen Graduate SchoolPeking UniversityShenzhenChina school of Electrical and Electronic Engineering Nanyang Technological UniversitySingaporeSingapore

Generating a realistic person's image from one source pose conditioned on another different target pose is a promising computer vision *** previous mainstream methods mainly focus on exploring the transformation relationship between the keypoint-based source pose and the target pose,but rarely investigate the region-based human semantic *** current methods that adopt the parsing map neither consider the precise local pose-semantic matching issues nor the correspondence between two different *** this study,a Region Semantics-Assisted Generative Adversarial Network(RSA-GAN)is proposed for the pose-guided person image gen-eration *** particular,a regional pose-guided semantic fusion module is first devel-oped to solve the imprecise match issue between the semantic parsing map from a certain source image and the corresponding keypoints in the source *** well align the style of the human in the source image with the target pose,a pose correspondence guided style injection module is designed to learn the correspondence between the source pose and the target *** addition,one gated depth-wise convolutional cross-attention based style integration module is proposed to distribute the well-aligned coarse style information together with the precisely matched pose-guided semantic information to-wards the target *** experimental results indicate that the proposed RSA-GAN achieves a 23%reduction in LPIPS compared to the method without using the seman-tic maps and a 6.9%reduction in FID for the method with semantic maps,respectively,and also shows higher realistic qualitative results.

关键词： deep learning image processing

Channel estimation-based time-frequency neural network for post-equalization in underwater visible light communication

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

Chinese Optics Letters 2024年第6期22卷 116-122页

作者： Haoyu Zhang Li Yao Chaoxu Chen Yuan Wei Chao Shen Jianyang Shi Junwen Zhang Ziwei Li Nan Chi Key Laboratory for Information Science of Electromagnetic Waves(MoE) Department of Communication Science and EngineeringSchool of Information Science and TechnologyFudan UniversityShanghai 200433China Shanghai CIC of LEO Satellite communication Technology Fudan UniversityShanghai 200433China Shanghai ERC of LEO Satellite communication and Application Fudan UniversityShanghai 200433China

This Letter proposes a post-equalizer for underwater visible light communication(UVLC) systems that combines channel estimation and joint time-frequency analysis, named channel-estimation-based bandpass variable-order time-frequency network(CBV-TFNet). By utilizing a bandpass variable-order loss function with communication prior knowledge, CBVTFNet enhances communication performance and training stability. It enables lightweight implementation and faster convergence through a channel estimation-based mask. The superior performance of the proposed equalization method over Volterra and deep neural network(DNN)-based methods has been studied experimentally. Using bit-power loading discrete multitone (DMT) modulation, the proposed method achieves a transmission bitrate of 4.956 Gbps through a 1.2 m underwater channel utilizing only 38.15% of real multiplication calculations compared to the DNN equalizer and achieving a bitrate gain of440 Mbps and a significantly larger dynamic range over the LMS-Volterra equalizer. Results highlight CBV-TFNet's potential for future post-equalization in UVLC systems.

关键词： underwater visible light communication channel estimation post-equalizer neural network

SAM3D: zero-shot 3D object detection via the segment anything model

在线全文

学校读者我要写书评

暂无评论

Science China(Information Sciences) 2024年第4期67卷 331-332页

作者： Dingyuan ZHANG Dingkang LIANG Hongcheng YANG Zhikang ZOU Xiaoqing YE Zhe LIU Xiang BAI school of Artificial Intelligence and Automation Huazhong University of Science and Technology Baidu Inc. school of Electronic Information and communication Huazhong University of Science and Technology

In the past few years, foundation models have thrived and succeeded in linguistic and visual tasks, showing astonishing zero-shot and few-shot capabilities. Their advances encourage researchers and industries to extend the boundaries of what artificial intelligence can do and have shown some fantastic products(e.g., ChatGPT [1]) with the potential to change the world.

关键词：

Effect of antenna helicity on discharge characteristics of helicon plasma under a divergent magnetic field

在线全文

同方期刊数据库

学校读者我要写书评

暂无评论

Plasma Science and Technology 2024年第6期26卷 50-59页

作者：孙萌徐晓芳王陈文尹贤轶陈强张海宝 Laboratory of Plasma Physics and Materials School of Printing and Packaging EngineeringBeijing Institute of Graphic CommunicationBeijing 102600People’s Republic of China

The characteristics of the blue core phenomenon observed in a divergent magnetic field helicon plasma are investigated using two different helical antennas, namely right-handed and lefthanded helical antennas. The mode transition, discharge image, spatial profiles of plasma density and electron temperature are diagnosed using a Langmuir probe, a Nikon D90 camera,an intensified charge-coupled device camera and an optical emission spectrometer, *** results demonstrated that the blue core phenomenon appeared in the upstream region of the discharge tube at a fixed magnetic field under both helical antennas. However, it is more likely to appear in a right-handed helical antenna, in which the plasma density and ionization rate of the helicon plasma are higher. The spatial profiles of the plasma density and electron temperature are also different in both axial and radial directions for these two kinds of helical antenna. The wavelength calculated based on the dispersion relation of the bounded whistler wave is consistent with the order of magnitude of plasma length. It is proved that the helicon plasma is part of the wave mode discharge mechanism.

关键词： helicon plasma non-uniform magnetic field helical antenna blue core discharge mechanism

Enhancing Deep Learning Semantics:The Diffusion Sampling and Label-Driven Co-Attention Approach

在线全文

学校读者我要写书评

暂无评论

Computers, Materials & Continua 2024年第5期79卷 1939-1956页

作者： ChunhuaWang Wenqian Shang Tong Yi Haibin Zhu State Key Laboratory of Media Convergence and communication Communication University of ChinaBeijing100024China school of Computer and Cyber Sciences Communication University of ChinaBeijing100024China school of Computer Science and Engineering Guangxi Normal UniversityGuilin541004China Department of Computer Science Nipissing UniversityNorth BayON P1B 8L7Canada

The advent of self-attention mechanisms within Transformer models has significantly propelled the advancement of deep learning algorithms,yielding outstanding achievements across diverse ***,self-attention mechanisms falter when applied to datasets with intricate semantic content and extensive dependency *** response,this paper introduces a Diffusion Sampling and Label-Driven Co-attention Neural Network(DSLD),which adopts a diffusion sampling method to capture more comprehensive semantic information of the ***,themodel leverages the joint correlation information of labels and data to introduce the computation of text representation,correcting semantic representationbiases in thedata,andincreasing the accuracyof semantic ***,the model computes the corresponding classification results by synthesizing these rich data semantic *** on seven benchmark datasets show that our proposed model achieves competitive results compared to state-of-the-art methods.

关键词： Semantic representation sampling attention label-driven co-attention attention mechanisms

A survey on intelligence-endogenous network:Architecture and technologies for future 6G

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

Intelligent and Converged Networks 2024年第1期5卷 53-67页

作者： Lanlan Li the school of communication and Information Engineering Shanghai Technical Institute of Electronics&InformationShanghai 201141China

With the maturity of 5G technology and global commercialization,scholars in institutions and industrial circles began to research 6G *** important innovation of 6G technology is to integrate artificial intelligence(AI)technology and communication technology to build a highly endogenous intelligent communication *** paper investigates the process of AI technology introduced into the field of communication and reviews the use cases of the simulation and application of AI algorithms being discussed in 3GPP meetings in industry *** this research report,we first investigate the progress of AI technology in 5G network architecture and then discuss the requirements of endogenous intelligent 6G networks,which leads to the possible network *** work aims to provide enlightening guidance for subsequent research of intelligence-endogenous 6G network.

关键词： 5G 6G intelligence-endogenous artificial intelligence

Joint Optimization of Resource Allocation and Trajectory Based on User Trajectory for UAV-Assisted Backscatter communication System

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

China communications 2024年第2期21卷 197-209页

作者： Peizhong Xie Junjie Jiang Ting Li Yin Lu school of communication and Information Engineering Nanjing University of Posts and TelecommunicationsNanjing 210003China Key Lab of Broadband Wireless communication and Sensor Network Technology(Nanjing University of Posts and Telecommunications) Ministry of EducationNanjing 210003China

The Backscatter communication has gained widespread attention from academia and industry in recent years. In this paper, A method of resource allocation and trajectory optimization is proposed for UAV-assisted backscatter communication based on user trajectory. This paper will establish an optimization problem of jointly optimizing the UAV trajectories, UAV transmission power and BD scheduling based on the large-scale channel state signals estimated in advance of the known user trajectories, taking into account the constraints of BD data and working energy consumption, to maximize the energy efficiency of the system. The problem is a non-convex optimization problem in fractional form, and there is nonlinear coupling between optimization *** iterative algorithm is proposed based on Dinkelbach algorithm, block coordinate descent method and continuous convex optimization technology. First, the objective function is converted into a non-fractional programming problem based on Dinkelbach method,and then the block coordinate descent method is used to decompose the original complex problem into three independent sub-problems. Finally, the successive convex approximation method is used to solve the trajectory optimization sub-problem. The simulation results show that the proposed scheme and algorithm have obvious energy efficiency gains compared with the comparison scheme.

关键词： energy efficiency joint optimization UAV-assisted backscatter communication user trajectory