检索结果-内蒙古大学图书馆

IEEE Transactions on Green Communications and Networking 2024年

作者： Sun, Wei Li, Jingsi Guo, Pengxing He, Xiangyu Wang, Zimo Hu, Cong Liu, Weichen Hou, Weigang Guo, Lei School of Communications and Information Engineering Chongqing University of Posts and Telecommunications Chongqing400065 China Institute of Intelligent Communications and Network Security Chongqing University of Posts and Telecommunications Chongqing400065 China Key Laboratory of Big Data Intelligent Computing Chongqing University of Posts and Telecommunications Chongqing400065 China Guangxi Key Laboratory of Automatic Detecting Technology and Instruments Guilin University of Electronic Technology Guilin541004 China School of Computer Science and Engineering Nanyang Technological University Singapore

Optical networks-on-chips (ONoCs) have emerged as the most promising solution for on-chip interconnection networks, offering distinct advantages such as high bandwidth and low latency. However, traditional wired-based ONoCs often suffer from a high average number of hops, leading to challenges such as increased insertion loss, crosstalk noise, and energy consumption. Therefore, this paper presents a novel ONoC structure known as the metasurface ONoC (MSONoC), which integrates wired waveguide and free-space optical (FSO) transmission with the assistance of metasurfaces. Specifically, this work encompasses the design of a passive metasurface structure tailored for ONoCs, a virtual network cluster partitioning method, an optical metasurface access points select strategy, and a shortest-path routing algorithm. Performance verification is conducted using a modified Noxim simulator. The simulation results demonstrate that the proposed MSONoC offers advantages over traditional NoCs, radio frequency (RF)-based hybrid wired/wireless NoCs, and wired-based ONoCs in terms of throughput, delay, energy consumption, and signal-to-noise ratio under both synthetic traffic and real traffic models. These improvements proves the effectiveness of the MSONoC architecture in optimizing network performance, making it a promising solution for high-performance, energy-efficient communication in on-chip networks. © 2017 IEEE.

关键词： Network-on-chip

来源：评论

学校读者我要写书评

暂无评论

Multi-Granularity Open Intent Classification via Adaptive Granular-Ball Decision Boundary

arXiv

引用

arXiv 2024年

作者： Li, Yanhua Ouyang, Xiaocao Pan, Chaofan Zhang, Jie Zhao, Sen Xia, Shuyin Yang, Xin Wang, Guoyin Li, Tianrui School of Computing and Artificial Intelligence Southwestern University of Finance and Economics Chengdu611130 China Engineering Research Center of Intelligent Finance Ministry of Education Chengdu611130 China Key Laboratory of Big Data Intelligent Computing Chongqing Key Laboratory of Computational Intelligence Chongqing University of Posts and Telecommunications Chongqing400065 China National Center for Applied Mathematics in Chongqing Chongqing Normal University Chongqing401331 China School of Computing and Artificial Intelligence Southwest Jiaotong University Chengdu611756 China

Open intent classification is critical for the development of dialogue systems, aiming to accurately classify known intents into their corresponding classes while identifying unknown intents. Prior boundary-based methods assumed known intents fit within compact spherical regions, focusing on coarse-grained representation and precise spherical decision boundaries. However, these assumptions are often violated in practical scenarios, making it difficult to distinguish known intent classes from unknowns using a single spherical boundary. To tackle these issues, we propose a Multi-granularity Open intent classification method via adaptive Granular-Ball decision boundary (MOGB). Our MOGB method consists of two modules: representation learning and decision boundary acquiring. To effectively represent the intent distribution, we design a hierarchical representation learning method. This involves iteratively alternating between adaptive granular-ball clustering and nearest sub-centroid classification to capture fine-grained semantic structures within known intent classes. Furthermore, multi-granularity decision boundaries are constructed for open intent classification by employing granular-balls with varying centroids and radii. Extensive experiments conducted on three public datasets demonstrate the effectiveness of our proposed method. Copyright © 2024, The Authors. All rights reserved.

关键词： Semantics

来源：评论

学校读者我要写书评

暂无评论

Symptom Dimensions and Cognitive Impairments in Individuals at Clinical High Risk for Psychosis

引用

Biological Psychiatry: Cognitive Neuroscience and Neuroimaging 2025年第6期10卷 646-655页

作者： Zhang, TianHong Xu, LiHua Wei, YanYan Cui, HuiRu Tang, XiaoChen Hu, YeGang Liu, HaiChun Wang, ZiXuan Chen, Tao Yi, ZhengHui Li, ChunBo Wang, JiJun Shanghai Mental Health Center Shanghai Jiaotong University School of Medicine Shanghai Engineering Research Center of Intelligent Psychological Evaluation and Intervention Shanghai Key Laboratory of Psychotic Disorders Shanghai China Department of Automation Shanghai Jiao Tong University Shanghai China Department of Psychology Shanghai Xinlianxin Psychological Counseling Center Shanghai China Big Data Research Laboratory University of Waterloo Waterloo ON Canada Labor and Worklife Program Harvard University Cambridge Massachusetts United States Center for Excellence in Brain Science and Intelligence Technology Chinese Academy of Sciences Shanghai China Institute of Psychology and Behavioral Science Shanghai Jiao Tong University Shanghai China

Background: Understanding the intricate relationships between symptom dimensions, clusters, and cognitive impairments is crucial for early detection and intervention in individuals at clinical high risk for psychosis. This study delves into this complex interplay in a clinical high risk sample with the aim of predicting the conversion to psychosis. Methods: A comprehensive cognitive assessment was performed in 744 clinical high risk individuals. The study included a 3-year follow-up period to allow assessment of conversion to psychosis. Symptom profiles were determined using the Structured Interview for Prodromal Syndromes. By applying factor analysis, symptom dimensions were categorized as dominant negative symptoms (NS), positive symptoms-stressful, and positive symptoms-odd. The factor scores were used to define 3 dominant symptom groups. Latent class analysis (LCA) and the factor mixture model (FMM) were employed to identify discrete clusters based on symptom patterns. The 3-class solution was chosen for the LCA and FMM analysis. Results: Individuals in the dominant NS group exhibited significantly higher conversion rates to psychosis than those in the other groups. Specific cognitive variables, including performance on the Brief Visuospatial Memory Test–Revised (odds ratio = 0.702, p = .001) and Neuropsychological Assessment Battery Mazes Test (odds ratio = 0.776, p = .024), significantly predicted conversion to psychosis. Notably, cognitive impairments associated with NS and positive symptoms-stressful groups affected different cognitive domains. LCA and FMM cluster 1, which was characterized by severe NS and positive symptoms-odd, exhibited more impairments in cognitive domains than other clusters. No significant difference in the conversion rate was observed among the LCA and FMM clusters. Conclusions: These findings highlight the importance of NS in the development of psychosis and suggest specific cognitive domains that are affected by symptom dimensions.

关键词： Negative symptoms Neurocognition Subtype Transition Ultra high risk

来源：评论

学校读者我要写书评

暂无评论

Diffusion-based Dynamic Contract for Federated AI Agent Construction in Mobile Metaverses

arXiv

引用

arXiv 2025年

作者： Wen, Jinbo Kang, Jiawen Zhang, Yang Zhong, Yue Niyato, Dusit Xu, Jie Tang, Jianhang Yuen, Chau College of Computer Science and Technology Nanjing University of Aeronautics and Astronautics China School of Automation Guangdong University of Technology China College of Computing and Data Science Nanyang Technological University Singapore Guangdong Provincial Key Laboratory of Future Networks of Intelligence The Chinese University of Hong Kong Shenzhen China State Key Laboratory of Public Big Data Guizhou University China School of Electrical and Electronics Engineering Nanyang Technological University Singapore

Mobile metaverses have attracted significant attention from both academia and industry, which are envisioned as the next-generation Internet, providing users with immersive and ubiquitous metaverse services through mobile devices. Driven by Large Language Models (LLMs) and Vision-Language Models (VLMs), Artificial Intelligence (AI) agents hold the potential to empower the creation, maintenance, and evolution of mobile metaverses. Currently, AI agents are primarily constructed using cloud-based LLMs and VLMs. However, several challenges hinder their effective implementation, including high service latency and potential sensitive data leakage during perception and processing. In this paper, we develop an edge-cloud collaboration-based federated AI agent construction framework in mobile metaverses. Specifically, Edge Servers (ESs), acting as agent infrastructures, collaboratively create agent modules in a distributed manner. The cloud server then integrates these modules into AI agents and deploys them at the edge, thereby enabling low-latency AI agent services for users. Considering that ESs may exhibit dynamic levels of willingness to participate in federated AI agent construction, we design a two-period dynamic contract model to continuously motivate ESs to participate in agent module creation, effectively addressing the dynamic information asymmetry between the cloud server and the ESs. Furthermore, we propose an Enhanced Diffusion Model-based Soft Actor-Critic (EDMSAC) algorithm to efficiently generate optimal dynamic contracts, in which dynamic structured pruning is applied to DM-based actor networks to enhance denoising efficiency and policy learning performance. Extensive simulations demonstrate the effectiveness and superiority of the EDMSAC algorithm and the proposed contract model. Copyright © 2025, The Authors. All rights reserved.

关键词： Reinforcement learning

来源：评论

学校读者我要写书评

暂无评论

Rgmef: A Lightweight Multi-Exposure Image Fusion Network Based on Retinex and Gradient Guidance

SSRN

引用

SSRN 2024年

作者： Feng, Yunfei Zhang, Mei Zhu, Jinhui School of Automation Science and Engineering South China University of Technology Guangzhou510641 China School of Software Engineering South China University of Technology Guangzhou510641 China Key Laboratory of Big Data and Intelligent Robot South China University of Technology Ministry of Education Guangzhou510641 China

The goal of the multi-exposure image fusion (MEF) is to create a high dynamic range (HDR) image with balanced exposures and more details by extracting and enhancing valid information from multiple low dynamic range (LDR) images with varied exposures. In recent years, deep learning has provided many solutions for MEF, and they have achieved remarkable effects. However, because of the high network complexity, these approaches often have low efficiency and limit the potential improvement of the effect. In order to fuse valuable information from different images, we introduce a lightweight but robust MEF network, which focus on the image internal characteristics rather than long-distance information. In this study, we first design a decomposed network inspired by Retinex theory to extract the illumination map from an image and obtain the gradient map through Sobel operator. Subsequently, we introduce two lightweight feature extraction modules to extract image features with the guidance of illumination and gradient maps respectively, and then fuse these features to produce an HDR result through three fusion blocks. Furthermore, we design a loss function to obtain a more suitable brightness range for the current scene. Numerous comparative experiments demonstrate the superiority of our method in terms of image texture, pixel intensity, and color retention. Moreover, some ablation studies indicate that our approach has a significant adaptability to source images. © 2024, The Authors. All rights reserved.

关键词： Image fusion

来源：评论

学校读者我要写书评

暂无评论

ISMI-VAE: A deep learning model for classifying disease cells using gene expression and SNV data

引用

Computers in Biology and Medicine 2024年 175卷 108485-108485页

作者： Li, Han Zhou, Yitao Zhao, Ningyuan Wang, Ying Lai, Yongxuan Zeng, Feng Yang, Fan Department of Automation Xiamen University Xiamen China National Institute for Data Science in Health and Medicine Xiamen University Xiamen361005 China Xiamen Key Laboratory of Big Data Intelligent Analysis and Decision Making Xiamen university Xiamen361000 China School of Informatics Xiamen University Xiamen China State Key Laboratory of Cellular Stress Biology School of Life Sciences Xiamen University China Research Unit of Cellular Stress of CAMS Cancer Research Center School of Medicine Xiamen University China

Various studies have linked several diseases, including cancer and COVID-19, to single nucleotide variations (SNV). Although single-cell RNA sequencing (scRNA-seq) technology can provide SNV and gene expression data, few studies have integrated and analyzed these multimodal data. To address this issue, we introduce Interpretable Single-cell Multimodal data Integration Based on Variational Autoencoder (ISMI-VAE). ISMI-VAE leverages latent variable models that utilize the characteristics of SNV and gene expression data to overcome high noise levels and uses deep learning techniques to integrate multimodal information, map them to a low-dimensional space, and classify disease cells. Moreover, ISMI-VAE introduces an attention mechanism to reflect feature importance and analyze genetic features that could potentially cause disease. Experimental results on three cancer data sets and one COVID-19 data set demonstrate that ISMI-VAE surpasses the baseline method in terms of both effectiveness and interpretability and can effectively identify disease-causing gene features. © 2024 Elsevier Ltd

关键词： Classification (of information)

来源：评论

学校读者我要写书评

暂无评论

Stealthy False data Injection Attacks in Multi-Channel Vehicular Communication: White-Box and Gray-Box Strategies

引用

IEEE Transactions on Vehicular Technology 2025年

作者： Xu, Chongcong Zheng, Fangfang Guo, Ge Southwest Jiaotong University School of Transportation and Logistics Chengdu611756 China Southwest Jiaotong University National Engineering Laboratory of Integrated Transportation Big Data Application Technology National United Engineering Laboratory of Integrated and Intelligent Transportation Chengdu611756 China Northeastern University State Key Laboratory of Synthetical Automation for Process Industries Shenyang110819 China Northeastern University at Qinhuangdao School of Control Engineering Qinhuangdao066004 China

Understanding the vulnerabilities of vehicular networks from an attacker's perspective is crucial for identifying stealthy false data injection (FDI) attacks and developing robust defense methods. This paper investigates stealthy FDI attacks in multi-channel vehicular communication systems. We construct a multi-source state estimation system using a sequential Kalman filter to integrate and estimate redundant state information transmitted by preceding vehicles across multiple channels. To enhance the system's resilience against FDI attacks, we incorporate a Sage-Husa adaptive filter and a chi-squared detector, which dynamically estimate and detect channel noise based on estimation errors. Within this framework, we utilize real-time state information and channel transmission residual to determine the probability of an attack window occurring. The attack process is divided into three phases, during which critical noise disturbances are introduced to create attack windows, allowing the attack to progressively align with the target state. Based on this analysis, we develop an optimization model for attack generation with complete channel information (AG-with-CCI) to facilitate a white-box attack. Additionally, acknowledging that attackers may not have access to real-time information from all channels, we analyze the iterative equations of state estimation errors and residuals to establish the conditions under which stealthy FDI attacks can be successfully executed, enabling the attack generation with attacked channel information (AG-with-ACI) to achieve a gray-box attack. Numerical simulations demonstrate that both attack methods can evade current detection mechanisms and achieve their intended objectives. © 2025 IEEE.

关键词： Adaptive filtering

来源：评论

学校读者我要写书评

暂无评论

CA-Edit: Causality-Aware Condition Adapter for High-Fidelity Local Facial Attribute Editing

arXiv

引用

arXiv 2024年

作者： Xian, Xiaole He, Xilin Niu, Zenghao Zhang, Junliang Xie, Weicheng Song, Siyang Yu, Zitong Shen, Linlin Computer Vision Institute School of Computer Science & Software Engineering Shenzhen University China National Engineering Laboratory for Big Data System Computing Technology Shenzhen University China Guangdong Key Laboratory of Intelligent Information Processing China University of Exeter United Kingdom Great Bay University China

For efficient and high-fidelity local facial attribute editing, most existing editing methods either require additional finetuning for different editing effects or tend to affect beyond the editing regions. Alternatively, inpainting methods can edit the target image region while preserving external areas. However, current inpainting methods still suffer from the generation misalignment with facial attributes description and the loss of facial skin details. To address these challenges, (i) a novel data utilization strategy is introduced to construct datasets consisting of attribute-text-image triples from a data-driven perspective, (ii) a Causality-Aware Condition Adapter is proposed to enhance the contextual causality modeling of specific details, which encodes the skin details from the original image while preventing conflicts between these cues and textual conditions. In addition, a Skin Transition Frequency Guidance technique is introduced for the local modeling of contextual causality via sampling guidance driven by low-frequency alignment. Extensive quantitative and qualitative experiments demonstrate the effectiveness of our method in boosting both fidelity and editability for localized attribute editing. The code is available at https://***/connorxian/CA-Edit. © 2024, CC BY.

关键词：

来源：评论

学校读者我要写书评

暂无评论

GM-DF: Generalized Multi-Scenario Deepfake Detection

arXiv

引用

arXiv 2024年

作者： Lai, Yingxin Yu, Zitong Yang, Jing Li, Bin Kang, Xiangui Shen, Linlin The School of Computing and Information Technology Great Bay University Dongguan523000 China National Engineering Laboratory for Big Data System Computing Technology Shenzhen University Shenzhen518060 China Shenzhen Institute of Artificial Intelligence and Robotics for Society Shenzhen University Shenzhen518060 China The Guangdong Key Laboratory of Information Security The School of Computer Science and Engineering Sun Yat-sen University Guangzhou510080 China Computer Vision Institute School of Computer Science & Software Engineering Shenzhen Institute of Artificial Intelligence and Robotics for Society Guangdong Key Laboratory of Intelligent Information Processing National Engineering Laboratory for Big Data System Computing Technology Shenzhen University Shenzhen518060 China

Existing face forgery detection usually follows the paradigm of training models in a single domain, which leads to limited generalization capacity when unseen scenarios and unknown attacks occur. In this paper, we elaborately investigate the generalization capacity of deepfake detection models when jointly trained on multiple face forgery detection datasets. We first find a rapid degradation of detection accuracy when models are directly trained on combined datasets due to the discrepancy across collection scenarios and generation methods. To address the above issue, a Generalized Multi-Scenario Deepfake Detection framework (GM-DF) is proposed to serve multiple real-world scenarios by a unified model. First, we propose a hybrid expert modeling approach for domain-specific real/forgery feature extraction. Besides, as for the commonality representation, we use CLIP to extract the common features for better aligning visual and textual features across domains. Meanwhile, we introduce a masked image reconstruction mechanism to force models to capture rich forged details. Finally, we supervise the models via a domain-aware meta-learning strategy to further enhance their generalization capacities. Specifically, we design a novel domain alignment loss to strongly align the distributions of the meta-test domains and meta-train domains. Thus, the updated models are able to represent both specific and common real/forgery features across multiple datasets. In consideration of the lack of study of multi-dataset training, we establish a new benchmark leveraging multi-source data to fairly evaluate the models' generalization capacity on unseen scenarios. Both qualitative and quantitative experiments on five datasets conducted on traditional protocols as well as the proposed benchmark demonstrate the effectiveness of our approach. The codes will be available on https://***/laiyingxin2/GM-DF. Copyright © 2024, The Authors. All rights reserved.

关键词： Image reconstruction

来源：评论

学校读者我要写书评

暂无评论

Padding Adversarial Training: Enhancing Robustness with Adversarial Space Hypothesis

SSRN

引用

SSRN 2022年

作者： Sun, Jiaze Wen, Sulei School of Computer Science and Technology Xi’an University of Posts and Telecommunications Xi’an710121 China Shaanxi Key Laboratory of Network Data Analysis and Intelligent Processing Xi’an710121 China Xi 'an Key Laboratory of Big Data and Intelligent Computing Xi’an710127 China

Adversarial training, represented by the projected gradient descent(PGD) adversarial training, effectively improves adversarial robustness within the upper bound of perturbation to the deep neural network(DNN) by applying adversarial examples to model retraining. When the perturbation is beyond the pale, though, the defense it provides to the model is invalid. Furthermore, it is also vulnerable to adaptive attacks and novel attacks uninvolved in adversarial training. To solve these problems, we propose the adversarial space hypothesis based on the manifold theory to explain the existence of adversarial examples from a feature space perspective. We further introduce padding adversarial training, a technique enhancing adversarial training with the adversarial space hypothesis. Padding adversarial training improves the generalization of the standard adversarial training’s robustness by 20% and provides an additional detection defense for adversarial examples with high perturbation. Moreover, we propose trap smoothing, which optimizes the target manifold’s distribution and establishes the inducing relationship between the target manifold and the feature space labelled by the padding classes. On MNIST and CIFAR-10, padding adversarial training is robust to white, black and adaptive attacks in low perturbation conditions and provides the 80%~98% detection rate of adversarial examples in high perturbation conditions. The code is available at:https://***/Ccsk-Xian/Padding-adversarial-training. © 2022, The Authors. All rights reserved.

关键词： Deep neural networks

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：