检索结果-内蒙古大学图书馆

Proceedings of the 23rd USENIX Conference on File and Storage Technologies

作者： Bing Tian Haikun Liu Yuhang Tang Shihai Xiao Zhuohui Duan Xiaofei Liao Hai Jin Xuecang Zhang Junhua Zhu Yu Zhang National Engineering Research Center for Big Data Technology and System Service Computing Technology and System Lab/Cluster and Grid Computing Lab School of Computer Science and Technology Huazhong University of Science and Technology China Huawei Technologies Co. Ltd Towards high-throughput and low-latency billion-scale vector search via CPU/GPU collaborative filtering and re-ranking

ISBN: (纸本)9781939133458

Approximate nearest neighbor search (ANNS) has emerged as a crucial component of database and AI infrastructure. Ever-increasing vector datasets pose significant challenges in terms of performance, cost, and accuracy for ANNS services. None of modern ANNS systems can address these issues simultaneously. In this paper, we present Fusion-ANNS, a high-throughput, low-latency, cost-efficient, and high-accuracy ANNS system for billion-scale datasets using SSDs and only one entry-level GPU. The key idea of Fusion-ANNS lies in CPU/GPU collaborative filtering and reranking mechanisms, which significantly reduce I/O operations across CPUs, GPU, and SSDs to break through the I/O performance bottleneck. Specifically, we propose three novel designs: (1) multi-tiered indexing to avoid data swapping between CPUs and GPU, (2) heuristic re-ranking to eliminate unnecessary I/Os and computations while guaranteeing high accuracy, and (3) redundant-aware I/O deduplication to further improve I/O efficiency. We implement FusionANNS and compare it with the state-of-the-art SSD-based ANNS system-SPANN and GPU-accelerated in-memory ANNS system-RUMMY. Experimental results show that FusionANNS achieves 1) 9.4-13.1× higher query per second (QPS) and 5.7-8.8× higher cost efficiency compared with SPANN; 2) and 2-4.9× higher QPS and 2.3-6.8× higher cost efficiency compared with RUMMY, while guaranteeing low latency and high accuracy.

关键词：

来源：评论

学校读者我要写书评

暂无评论

PriFFT: Privacy-preserving Federated Fine-tuning of Large Language Models via Function Secret Sharing

arXiv

引用

arXiv 2025年

作者： You, Zhichao Dong, Xuewen Cheng, Ke Mu, Xutong Fu, Jiaxuan Ma, Shiyang Qu, Qiang Shen, Yulong School of Computer Science & Technology Xidian University Shaanxi Key Laboratory of Network and System Security Xi’an China Shenzhen Institues of Advanced Technology Chinese Academy of Sciences China

Fine-tuning large language models (LLMs) raises privacy concerns due to the risk of exposing sensitive training data. Federated learning (FL) mitigates this risk by keeping training samples on local devices, but recent studies show that adversaries can still infer private information from model updates in FL. Additionally, LLM parameters are typically shared publicly during federated fine-tuning, while developers are often reluctant to disclose these parameters, posing further security challenges. Inspired by the above problems, we propose PriFFT, a privacy-preserving federated fine-tuning mechanism, to protect both the model updates and parameters. In PriFFT, clients and the server share model inputs and parameters by secret sharing, performing secure fine-tuning on shared values without accessing plaintext data. Due to considerable LLM parameters, privacy-preserving federated fine-tuning invokes complex secure calculations and requires substantial communication and computation resources. To optimize the efficiency of privacy-preserving federated fine-tuning of LLMs, we introduce function secret-sharing protocols for various operations, including reciprocal calculation, tensor products, natural exponentiation, softmax, hyperbolic tangent, and dropout. The proposed protocols achieve up to 4.02× speed improvement and reduce 7.19× communication overhead compared to the implementation based on existing secret sharing methods. Besides, PriFFT achieves a 2.23× speed improvement and reduces 4.08× communication overhead in privacy-preserving fine-tuning without accuracy drop compared to the existing secret sharing methods. Copyright © 2025, The Authors. All rights reserved.

关键词： Differential privacy

来源：评论

学校读者我要写书评

暂无评论

RWORT: A Read and Write Optimized Radix Tree for Persistent Memory

RWORT: A Read and Write Optimized Radix Tree for Persistent ...

引用

IEEE International Conference on computer Design: VLSI in computers and Processors, (ICCD)

作者： Jinlei Hu Zijie Wei Jianxi Chen Dan Feng Wuhan National Laboratory for Optoelectronics Key Laboratory of Information Storage System of MoE School of Computer Science and Technology Huazhong University of Science and Technology Wuhan China

Tree index structures are widely employed in modern storage systems to support high-performance queries. Persistent memory (PM) brings a new opportunity and challenge for tree indexes. Among persistent tree indexes, we find the radix tree is more suitable than B-Tree for the byte-ability of PM. However, the hierarchy of radix remains excessively high, resulting in high read latency. Node splitting imposes a significant overhead on PM. To address these challenges, we propose RWORT, a read and write optimized radix tree for PM. The key focus of RWORT is to minimize random access in PM and provide efficient write operations. RWORT proposed a hierarchical compression mechanism to significantly reduce the tree height. Additionally, RWORT incorporates mini bloom filters to reduce unnecessary access on PM. For efficient write operations, RWORT uses the lazy split flag and the double-linked pointers to reduce the critical path delay. Furthermore, RWORT introduces a low-overhead ring-based bit tree allocator that improves allocation efficiency on PM. Our experiments show that RWORT improves up to 1.62x/4.91x respectively compared to the state-of-the-art radix tree/B-Tree. RWORT also exhibits higher performance in real-world storage systems such as Memcached.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Detecting and Corrupting Convolution-based Unlearnable Examples 39

Detecting and Corrupting Convolution-based Unlearnable Examp...

引用

39th Annual AAAI Conference on Artificial Intelligence, AAAI 2025

作者： Li, Minghui Wang, Xianlong Yu, Zhifei Hu, Shengshan Zhou, Ziqi Zhang, Longling Zhang, Leo Yu School of Software Engineering Huazhong University of Science and Technology China Hubei Key Laboratory of Distributed System Security Hubei Engineering Research Center on Big Data Security China School of Cyber Science and Engineering Huazhong University of Science and Technology China School of Computer Science and Technology Huazhong University of Science and Technology China School of Information and Communication Technology Griffith University Australia

ISBN: (纸本)157735897X

Convolution-based unlearnable examples (UEs) employ class-wise multiplicative convolutional noise to training samples, severely compromising model performance. This fire-new type of UEs have successfully countered all defense mechanisms against UEs. The failure of such defenses can be attributed to the absence of norm constraints on convolutional noise, leading to severe blurring of image features. To address this, we first design an Edge Pixel-based Detector (EPD) to identify convolution-based UEs. Upon detection of them, we propose the first defense scheme against convolution-based UEs, COrrupting these samples via random matrix multiplication by employing bilinear INterpolation (COIN) such that disrupting the distribution of class-wise multiplicative noise. To evaluate the generalization of our proposed COIN, we newly design two convolution-based UEs called VUDA and HUDA to expand the scope of convolution-based UEs. Extensive experiments demonstrate the effectiveness of detection scheme EPD and that our defense COIN outperforms 11 state-of-the-art (SOTA) defenses, achieving a significant improvement on the CIFAR and ImageNet datasets. Copyright © 2025, Association for the Advancement of Artificial Intelligence (***). All rights reserved.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Adaptive Fake Audio Detection with Low-Rank Model Squeezing

Adaptive Fake Audio Detection with Low-Rank Model Squeezing

引用

2023 Workshop on Deepfake Audio Detection and Analysis, DADA 2023

作者： Zhang, Xiaohui Yi, Jiangyan Tao, Jianhua Wang, Chenglong Xu, Le Fu, Ruibo State Key Laboratory of Multimodal Artificial Intelligence System Institute of Automation Chinese Academy of Sciences China School of Computer and Information Technology Beijing Jiaotong University China University of Science and Technology of China China Department of Automation Tsinghua University China School of Artificial Intelligence University of Chinese Academy of Sciences China

The rapid advancement of spoofing algorithms necessitates the development of robust detection methods capable of accurately identifying emerging fake audio. Traditional approaches, such as finetuning on new datasets containing these novel spoofing algorithms, are computationally intensive and pose a risk of impairing the acquired knowledge of known fake audio types. To address these challenges, this paper proposes an innovative approach that mitigates the limitations associated with finetuning. We introduce the concept of training low-rank adaptation matrices tailored specifically to the newly emerging fake audio types. During the inference stage, these adaptation matrices are combined with the existing model to generate the final prediction output. Extensive experimentation is conducted to evaluate the efficacy of the proposed method. The results demonstrate that our approach effectively preserves the prediction accuracy of the existing model for known fake audio types. Furthermore, our approach offers several advantages, including reduced storage memory requirements and lower equal error rates compared to conventional finetuning methods, particularly on specific spoofing algorithms. © 2023 CEUR-WS. All rights reserved.

关键词： Fake detection

来源：评论

学校读者我要写书评

暂无评论

Multi-level graph contrastive learning for cold-start recommendation in mashup development

引用

Information sciences 2025年 717卷

作者： Sun, Mengmeng Xu, Yueshen Tan, Zeyu You, Dianlong Chen, Zhen School of Information Science and Engineering Yanshan University Qinhuangdao066004 China School of Computer Science and Technology Xidian University Xi'an710071 China Hebei Key Laboratory for Computer Virtual Technology and System Integration Yanshan University Qinhuangdao066004 China

One longstanding challenge in cloud API recommender systems is the Mashup cold-start problem, i.e., to recommend suitable cloud APIs for new Mashups without any historical invocation records. To address this challenge, researchers typically integrate various auxiliary information and feed them into a supervised learning model, or provide contextual information beyond interaction behaviors via explicit graph connections. However, these researches fail to 1) explore the functionally complementary relationship between Mashups and cloud APIs, and 2) model their semantic associations. To this end, we propose a Multi-level Graph Contrastive Learning, namely MGCL. Unlike traditional single-view methods, we comprehensively model Mashup-API relationships from three views: interaction, functionally complementary, and semantics. The interaction view captures dynamic behaviors, while the other two reveal associations based on static attributes. Additionally, we design a multi-level graph contrastive learning mechanism to fuse information from these views in a self-supervised manner, enabling both cold- and warm-start Mashup representations to benefit from dynamic interaction behaviors and static attributes. Extensive experiments on real-world datasets demonstrate that MGCL outperforms state-of-the-art methods in both Mashup cold- and warm-start scenarios. The implementations are available at: https://***/MengMeng3399/MGCL. © 2025 Elsevier Inc.

关键词： Contrastive Learning

来源：评论

学校读者我要写书评

暂无评论

Digital-Based Inverse Design for Ultra-Compact Power Splitter on LNOI

Digital-Based Inverse Design for Ultra-Compact Power Splitte...

引用

2023 Asia Communications and Photonics Conference/2023 International Photonics and Optoelectronics Meetings, ACP/POEM 2023

作者： Zhou, Lingjun Ma, Hansi Nie, Xiaomin Li, Yunchen He, Zhixue Wang, Lei Li, Ke Zhang, Fan Peng Cheng Laboratory Shenzhen China School of Electronic and Computer Engineering Peking University Shenzhen China Center of Material Science National University of Defense Technology Changsha China State Key Laboratory of Advanced Optical Communication System and Networks School of Electronics Peking University Beijing China

ISBN: (纸本)9798350312614

We propose and experimentally demonstrate an ultra-compact and efficient power splitter on LNOI. The device is realized through the inverse design of a digital meta-structure and occupies a small footprint of only 4.8×4.8μm2. The measured results show that the insertion losses are below 2.5dB across the wavelength band of 1500-1560 nm, and the power imbalance between the two output ports is within 0.3 dB. © 2023 IEEE.

关键词： inverse design LNOI power splitter

来源：评论

学校读者我要写书评

暂无评论

From laboratory to Field:Unsupervised Domain Adaptation for Plant Disease Recognition in the Wild

引用

Plant Phenomics 2023年第2期5卷 195-208页

作者： Xinlu Wu Xijian Fan Peng Luo Sruti Das Choudhury Tardi Tjahjadi Chunhua Hu College of Information Science and Technology Nanjing Forestry UniversityNanjing 210037China Institute of Forest Resource Information Techniques Chinese Academy of ForestryBeijing 100091China Key Laboratory of Forestry Remote Sensing and Information System National Forestry and Grassland AdministrationBeijing 100091China Department of Computer Science and Engineering University of Nebraska-LincolnLincolnNE 68588USA School of Engineering University of WarwickCoventry CV47ALUK

Plant disease recognition is of vital importance to monitor plant development and predicting crop ***,due to data degradation caused by different conditions of image acquisition,e.g.,laboratory *** environment,machine learning-based recognition models generated within a specific dataset(source domain)tend to lose their validity when generalized to a novel dataset(target domain).To this end,domain adaptation methods can be leveraged for the recognition by learning invariant representations across *** this paper,we aim at addressing the issues of domain shift existing in plant disease recognition and propose a novel unsupervised domain adaptation method via uncertainty regularization,namely,Multi-Representation Subdomain Adaptation Network with Uncertainty Regularization for Cross-Species Plant Disease Classification(MSUN).Our simple but effective MSUN makes a breakthrough in plant disease recognition in the wild by using a large amount of unlabeled data and via nonadversarial ***,MSUN comprises multirepresentation,subdomain adaptation modules and auxiliary uncertainty *** multirepresentation module enables MSUN to learn the overall structure of features and also focus on capturing more details by using the multiple representations of the source *** effectively alleviates the problem of large interdomain *** adaptation is used to capture discriminative properties by addressing the issue of higher interclass similarity and lower intraclass ***,the auxiliary uncertainty regularization effectively suppresses the uncertainty problem due to domain *** was experimentally validated to achieve optimal results on the PlantDoc,Plant-Pathology,Corn-Leaf-Diseases,and Tomato-Leaf-Diseases datasets,with accuracies of 56.06%,72.31%,96.78%,and 50.58%,respectively,surpassing other state-of-the-art domain adaptation techniques considerably.

关键词： Plant breakthrough details

来源：评论

学校读者我要写书评

暂无评论

CC-Net: Consistency Learning Combined with Contrastive Learning for Kidney Ultrasound Segmentation

CC-Net: Consistency Learning Combined with Contrastive Learn...

引用

IEEE International Conference on systems, Man and Cybernetics

作者： Yang Luo Jun Liu Mengqian Ding School of Computer Science and Technology Wuhan University of Science and Technology Wuhan China Hubei Province Key Laboratory of Intelligent Information Processing and Real-Time Industrial System Wuhan China

ISBN: (数字)9781665410205

ISBN: (纸本)9781665410212

Due to the scarcity of medical images with labels and the fact that medical image data is not easily accessible. Kidney ultrasound (KUS) images have speckle noise, low image quality, acoustic shadows, and other reasons, which pose significant challenges to the kidney ultrasound image segmentation task. To solve the above two problems, a novel consistency learning and contrastive learning-based kidney ultrasound image segmentation method (CC-Net) has been proposed. Specifically, the student-teacher model was adopted as the semi-supervised learning framework and uncertainty estimation was introduced. The Consensus Uncertainty Map (CUM) strategy was proposed to generate segmentation results for the teacher model when calculating uncertainty. To solve the problems of acoustic shadows and low image quality in kidney ultrasound images, a kidney ultrasound image data augmentation method (KUS-AUG) and a convolutional block (KUS-CONV) tailored to the characteristics of kidney ultrasound images were proposed. In addition, contrastive learning was implemented to enhance the encoder's ability to learn more semantic information, due to its inherent advantage in learning data distribution. Finally, a series of experiments were conducted on the ultrasound image dataset and the method achieved the best performance.

关键词： Image quality Image segmentation Ultrasonic imaging Uncertainty Noise Contrastive learning Semisupervised learning Acoustics Kidney Biomedical imaging

来源：评论

学校读者我要写书评

暂无评论

Tailoring Highly Confined Electric-field Based Optical Skyrmions

Tailoring Highly Confined Electric-field Based Optical Skyrm...

引用

Progress in Electromagnetic Research Symposium (PIERS)

作者： J. Chen X. Shen Y. Zeng Q. Zhan School of Optical-Electrical and Computer Engineering University of Shanghai for Science and Technology Shanghai China Shanghai Key Laboratory of Modern Optical System University of Shanghai for Science and Technology Shanghai China

ISBN: (数字)9798350375909

ISBN: (纸本)9798350375916

Optical skyrmions are quasi-particles with nontrivial topological textures, which hold potential applications in optical information processing, transmission, and storage. There are various three-dimensional physical vectors to construct the optical skyrmions, such as spin skyrmions, Stokes skyrmions, Poynting vector skyrmions, and so on. Here, we present a method to construct tightly focused electric-field skyrmions based on the array antenna radiation theory. A multiple concentric-ring array of dipoles (MCAD) is placed at the focal region of the high numerical aperture (NA) lens to mimic the topological structure of the desired electric field skyrmion, and the radiation pattern of the MCAD is derived to form the incident fields at the pupil planes of the high NA lenses in a 4π focusing setup. The characteristics of the generated focal field are quantitatively evaluated by using the Debye diffraction integral theory. The results demonstrate that two types of electric field vector topologies, i.e., skyrmion and meron, can be created with adjustable polarity, vorticity, and helicity. Additionally, the normal direction of the plane containing the skyrmion or meron can also be customized, increasing the degrees of freedom in tailoring optical skyrmions.

关键词： Electric potential Optical diffraction Information processing Vectors Topology Electric fields Pupils

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：