检索结果-内蒙古大学图书馆

arXiv 2025年

作者： Lu, Zhipeng Shenzhen MSU-BIT University Guangdong Laboratory of Machine Perception and Intelligent Computing Guangdong Shenzhen518172 China

We focus on establishing the foundational paradigm of a novel optimizationtheory based on convolution with convex kernels. Our goal is to devise amorally deterministic model of locating the global optima of an arbitraryfunction, which is distinguished from most commonly used statistical *** preliminary numerical results are provided to test the efficiency ofsome specific algorithms derived from our paradigm, which we hope to stimulatefurther practical interest. © 2025, CC BY.

关键词： Convolution

来源：评论

学校读者我要写书评

暂无评论

Speaker Extraction with Verification of Present and Absent Target Speakers

引用

Journal of Shanghai Jiaotong University (Science) 2025年 1-6页

作者： Zhang, Ke Borsdorf, Marvin Liu, Tianchi Wang, Shuai Wei, Yangjie Li, Haizhou Key Laboratory of Intelligent Computing in Medical Image Northeastern University Shenyang110819 China Shenzhen Guangdong518000 China Machine Listening Lab University of Bremen Bremen28359 Germany Department of Electrical and Computer Engineering National University of Singapore Singapore119077 Singapore

Target speaker extraction (TSE) models are expected to extract the target speech from a cocktail party mixture signal. When only trained with present target speaker samples (PT), these models output noise in the absence of the target speaker (AT). One may enhance the TSE quality by providing the information about the PT and AT. However, the detection of the target speaker is not perfect. In this paper, we propose a new model, TSEV, which performs target speaker extraction and speaker verification simultaneously. The TSEV model outputs an extracted speech and generates two speaker embeddings per inference to detect the target speaker. By sharing the speaker encoder and low-level modules, the speaker verification task can be performed in low signal-to-noise ratio scenarios. We train the TSEV model on multi-talker PT and AT conditions with fully overlapped speech. Experiments verify the superiority of jointly performing two tasks in the proposed model. The TSEV model achieves better verification performance without degrading the extraction performance compared with the baseline. © Shanghai Jiao Tong University 2025.

关键词： Speech recognition

来源：评论

学校读者我要写书评

暂无评论

Large-Scale Riemannian Meta-Optimization via Subspace Adaptation

arXiv

引用

arXiv 2025年

作者： Yu, Peilin Wu, Yuwei Gao, Zhi Fan, Xiaomeng Jia, Yunde Beijing Key Laboratory of Intelligent Information Technology School of Computer Science & Technology Beijing Institute of Technology China Guangdong Laboratory of Machine Perception and Intelligent Computing Shenzhen MSU-BIT University China

Riemannian meta-optimization provides a promising approach to solving non-linear constrained optimization problems, which trains neural networks as optimizers to perform optimization on Riemannian manifolds. However, existing Riemannian meta-optimization methods take up huge memory footprints in large-scale optimization settings, as the learned optimizer can only adapt gradients of a fixed size and thus cannot be shared across different Riemannian parameters. In this paper, we propose an efficient Riemannian meta-optimization method that significantly reduces the memory burden for large-scale optimization via a subspace adaptation scheme. Our method trains neural networks to individually adapt the row and column subspaces of Riemannian gradients, instead of directly adapting the full gradient matrices in existing Riemannian meta-optimization methods. In this case, our learned optimizer can be shared across Riemannian parameters with different sizes. Our method reduces the model memory consumption by six orders of magnitude when optimizing an orthogonal mainstream deep neural network (e.g. ResNet50). Experiments on multiple Riemannian tasks show that our method can not only reduce the memory consumption but also improve the performance of Riemannian meta-optimization. © 2025, CC BY-NC-ND.

关键词： Constrained optimization

来源：评论

学校读者我要写书评

暂无评论

Exploiting Robust Model Watermarking Against the Model Fine-Tuning Attack via Flat Minima Aware Optimizers

Exploiting Robust Model Watermarking Against the Model Fine-...

引用

International Conference on Acoustics, Speech, and Signal Processing (ICASSP)

作者： Dongdong Lin Yue Li Bin Li Jiwu Huang Guangdong Provincial Key Laboratory of Intelligent Information Processing Shenzhen University Shenzhen China Shenzhen Key Laboratory of Media Security Shenzhen University Shenzhen China SZU-AFS Joint Innovation Center for AI Technology Shenzhen University Shenzhen China Xiamen Key Laboratory of Data Security and Blockchain Technology Huaqiao University Xiamen China Guangdong Laboratory of Machine Perception and Intelligent Computing Shenzhen MSU-BIT University Shenzhen China

ISBN: (数字)9798350368741

ISBN: (纸本)9798350368758

With the rapid advancement of deep neural networks (DNNs), model watermarking has emerged as a widely adopted technique for safeguarding model copyrights. A prevalent method involves utilizing a watermark decoder to retrieve watermark bits from generated outputs, but such methods are often vulnerable to model fine-tuning attacks. Traditionally, this challenge is mitigated through adversarial training or data augmentation, both of which significantly increase the computational burden. In this paper, we present a solution employing Flat Minima Aware (FMA) optimizers to bolster the robustness of model watermarking without requiring additional training data. By optimizing the watermark loss with flat minima awareness, our approaches significantly enhance the robustness of watermarks against the model fine-tuning attack. Comprehensive experiments have demonstrated our method’s superior ability to preserve watermark integrity. These findings suggest that this innovative optimization strategy offers a robust and efficient pathway for protecting models, thereby contributing to more secure and reliable model copyright protection mechanisms.

关键词： Training Resistance Computational modeling Training data Watermarking Signal processing Diffusion models Robustness Speech processing Optimization

来源：评论

学校读者我要写书评

暂无评论

Accurate and Efficient Privacy-Preserving Feature Extraction on Encrypted Images

引用

IEEE Transactions on Dependable and Secure computing 2025年

作者： Zheng, Peijia Fang, Xiongjie Zhou, Zhipeng Yang, Rui Lu, Wei Cao, Xiaochun Huang, Jiwu Sun Yat-Sen University School of Computer Science and Engineering Guangdong Province Key Laboratory of Information Security Technology MoE Key Laboratory of Information Technology Guangzhou510006 China Alibaba Group Hanzhou311121 China Shenzhen Campus of Sun Yat-sen University School of Cyber Science and Technology Shenzhen518100 China Shenzhen MSU-BIT University Guangdong Laboratory of Machine Perception and Intelligent Computing Faculty of Engineering Shenzhen518116 China

In cloud computing, it is necessary to outsource image processing algorithms securely without exposing private image content. The scale-invariant feature transform (SIFT) is a famous local descriptor widely used in computer vision. There are already some privacy-preserving schemes for computing SIFT on encrypted images. However, the state-of-the-art works have to convert fixed-point numbers into their binary representations, which reduces efficiency and accuracy. In this paper, we propose a novel privacy-preserving SIFT scheme built from secure protocols designed explicitly for fixed-point numbers to solve this problem. Specifically, using RLWE-based homomorphic encryption, we propose word-wise protocols to perform secure division, square root operation, comparison, derivation, and matrix inversion in a single-instruction multiple-data manner. These protocols allow direct processing of fixed-point numbers without converting them to binary numbers, thus achieving high computational efficiency. We have also realized critical SIFT steps missing from previous works, including Euclidean gradient amplitude computation, histogram peak interpolation, and precise interval localization, leading to improved accuracy of SIFT features in the encrypted domain. We conduct security analysis and perform extensive experiments to evaluate the execution efficiency and accuracy. The experimental results show that the proposed scheme outperforms the state-of-the-art works in terms of computational efficiency and accuracy. © 2004-2012 IEEE.

关键词： Differential privacy

来源：评论

学校读者我要写书评

暂无评论

End-to-End Open-Vocabulary Video Visual Relationship Detection Using Multi-Modal Prompting

引用

IEEE Transactions on Pattern Analysis and machine Intelligence 2025年

作者： Wang, Yongqi Wu, Xinxiao Yang, Shuo Luo, Jiebo Beijing Institute of Technology Beijing Key Laboratory of Intelligent Information Technology School of Computer Science and Technology Beijing100081 China Shenzhen MSU-BIT University Guangdong Provincial Laboratory of Machine Perception and Intelligent Computing Shenzhen518172 China University of Rochester Department of Computer Science RochesterNY14627 United States

Open-vocabulary video visual relationship detection aims to expand video visual relationship detection beyond annotated categories by detecting unseen relationships between both seen and unseen objects in videos. Existing methods usually use trajectory detectors trained on closed datasets to detect object trajectories, and then feed these trajectories into large-scale pre-trained vision-language models to achieve open-vocabulary classification. Such heavy dependence on the pre-trained trajectory detectors limits their ability to generalize to novel object categories, leading to performance degradation. To address this challenge, we propose to unify object trajectory detection and relationship classification into an end-to-end open-vocabulary framework. Under this framework, we propose a relationship-aware open-vocabulary trajectory detector. It primarily consists of a query-based Transformer decoder, where the visual encoder of CLIP is distilled for frame- wise open-vocabulary object detection, and a trajectory associator. To exploit relationship context during trajectory detection, a relationship query is embedded into the Transformer decoder, and accordingly, an auxiliary relationship loss is designed to enable the decoder to perceive the relationships between objects explicitly. Moreover, we propose an open-vocabulary relationship classifier that leverages the rich semantic knowledge of CLIP to discover novel relationships. To adapt CLIP well to relationship classification, we design a multi-modal prompting method that employs spatio-temporal visual prompting for visual representation and vision-guided language prompting for language input. Extensive experiments on two public datasets, VidVRD and VidOR, demonstrate the effectiveness of our framework. Our framework is also applied to a more difficult cross-dataset scenario to further demonstrate its generalization ability. The code for this paper is available at https://***/wangyongqi558/EOV-MMP-VidVRD. © 1

关键词： Object detection

来源：评论

学校读者我要写书评

暂无评论

Block-diagonal structure learning for subspace clustering

引用

Expert Systems with Applications 2025年 285卷

作者： Xing, Zheng Zhao, Weibing College of Computer Science and Software Engineering Shenzhen University Guangdong Shenzhen518060 China Guangdong Laboratory of Machine Perception and Intelligent Computing Shenzhen MSU-BIT University Guangdong Shenzhen518172 China

Subspace clustering plays a pivotal role in discovering underlying structures within high-dimensional data by identifying clusters in lower-dimensional subspaces. However, existing works devote less attention to the potential contribution of the inherent structure to both similarity matrix construction and clustering structure identification. This paper aims to advance subspace clustering by investigating the block-diagonal structure of the similarity matrix. The core idea involves constructing a similarity matrix that assesses the similarities among high-dimensional data points, followed by a cluster-ordering procedure that establishes the necessary permutation. This facilitates the direct identification of clustering structures through the recognition of diagonal blocks in the permuted similarity matrix. Consequently, the main challenges involve permuting the similarity matrix to achieve the desired structure and efficiently identifying the diagonal blocks within the permuted similarity matrix. To address these challenges, we develop a traversal algorithm that detects high-density clusters in the dataset and generates an enhanced cluster ordering. The block-diagonal structure is achieved through permutation aligned with the traversal sequence. Additionally, we introduce a novel greedy segmentation algorithm to autonomously identify all diagonal blocks within the permuted similarity matrix, offering theoretical optimality. Experimental results demonstrate that the proposed method enhances the performance of state-of-the-art subspace clustering methods on synthetic and real datasets. © 2025 Elsevier Ltd

关键词： Clustering algorithms

来源：评论

学校读者我要写书评

暂无评论

Real-world Cyber Security Demonstration for Networked Electric Drives

引用

IEEE Journal of Emerging and Selected Topics in Power Electronics 2025年

作者： Yang, He Yang, Bowen Coshatt, Stephen Li, Qi Hu, Kun Hammond, Bryan Cooper Ye, Jin Parasuraman, Ramviyas Song, Wenzhan The Intelligent Power Electronics and Electric Machine Laboratory University of Georgia AthensGA30602 United States The Power Systems Engineering Center National Renewable Energy Laboratory GoldenCO80401 United States The Center for Cyber-Physical Systems University of Georgia AthensGA30602 United States Eaton ClevelandOH44122 United States Sino-German College of Intelligent Manufacturing Shenzhen Technology University Guangdong Shenzhen518118 China Research Lab the School of Computing University of Georgia AthensGA30602 United States

In this paper, we present the design and implementation of a cyber-physical security testbed for networked electric drive systems, aimed at conducting real-world security demonstrations. To our knowledge, this is one of the first security testbeds for networked electric drives, seamlessly integrating the domains of power electronics and computer science, and cybersecurity. By doing so, the testbed offers a comprehensive platform to explore and understand the intricate and often complex interactions between cyber and physical systems. The core of our testbed consists of four electric machine drives, meticulously configured to emulate small-scale but realistic information technology (IT) and operational technology (OT) networks. This setup both provides a controlled environment for simulating a wide array of cyber attacks, and mirrors potential real-world attack scenarios with a high degree of fidelity. The testbed serves as an invaluable resource for the study of cyberphysical security, offering a practical and dynamic platform for testing and validating cybersecurity measures in the context of networked electric drive systems. As a concrete example of the testbed’s capabilities, we have developed and implemented a Python-based script designed to execute step-stone attacks over a wireless local area network (WLAN). This script leverages a sequence of target IP addresses, simulating a real-world attack vector that could be exploited by adversaries. To counteract such threats, we demonstrate the efficacy of our developed cyberattack detection algorithms, which are integral to our testbed’s security framework. Furthermore, the testbed incorporates a real-time visualization system using InfluxDB and Grafana, providing a dynamic and interactive representation of networked electric drives and their associated security monitoring mechanisms. This visualization component not only enhances the testbed’s usability but also offers insightful, real-time data for researchers and

关键词： Testbeds

来源：评论

学校读者我要写书评

暂无评论

2-D shape completion with shape priors

引用

Chinese Science Bulletin 2013年第27期58卷 3430-3436页

作者： GUO Ge JIANG TingTing WANG YiZhou GAO Wen Key Laboratory of Intelligent Information Processing Institute of Computing Technology Chinese Academy of Sciences Graduate University of Chinese Academy of Sciences National Engineering Laboratory for Video Technology Key Laboratory of Machine Perception School of EECS Peking University

We propose new techniques for 2-D shape/contour completion, which is one of the important research topics related to shape analysis and computer vision, e.g. the detection of incomplete objects due to occlusion and noises. The purpose of shape completion is to find the optimal curve segments that fill the missing contour parts, so as to acquire the best estimation of the original complete object shapes. Unlike the previous work using local smoothness or minimum curvature priors, we solve the problem under a Bayesian formulation taking advantage of global shape prior knowledge. With the priors, our methods are expert in recovering significant shape structures and dealing with large occlusion cases. There are two different priors adopted in this paper: (i) A generic prior model that prefers minimal global shape transformation (including non-rigid deformation and affine transformation with respect to a reference object shape) of the recovered complete shape; and (ii) a class-specific shape prior model learned from training examples of an object category, which prefers the reconstructed shape to follow the learned shape variation models of the category. Efficient contour completion algorithms are suggested corresponding to the two types of priors. Our experimental results demonstrate the advantage of the proposed shape completion approaches compared to the existing techniques, especially for objects with complex structure under severe occlusion.

关键词：形状结构计算机视觉贝叶斯公式电子相关最佳估计最小曲率先验知识仿射变换

来源：评论

学校读者我要写书评

暂无评论

A customized CNN architecture for detecting double and multiple JPEG compression in small blocks

引用

Engineering Analysis with Boundary Elements 2025年 178卷

作者： Israr Hussain Naeem Hussain Shunquan Tan College of Computer Science and Software Engineering Shenzhen University Shenzhen 518060 China College of Electronic and Information Engineering Shenzhen 518060 China School of Information and Communications Technology Shenzhen City Polytechnic Shenzhen 518116 China Guangdong Laboratory of Machine Perception and Intelligent Computing Faculty of Engineering Shenzhen MSU-BIT University Shenzhen 518116 China

Detecting double JPEG compression is crucial for verifying the authenticity and integrity of JPEG images. However, this poses significant challenges, particularly when the quality factor (QF) of the first compression is substantially higher than of the second compression, and when working with small-sized blocks. To overcome these challenges, we propose a customized convolutional neural network (CNN) architecture specifically designed for detecting double and multiple JPEG compression. The main contribution of our work is the innovative use of raw DCT coefficients as input to the CNN, enabling end-to-end feature extraction and enhancing the detection of subtle compression artifacts that might be missed by traditional techniques. The framework incorporates an ABS layer after the first convolution to eliminate the symmetry information that are not useful for JPEG compression detection, applies batch normalization (BN) to stabilize training, and reduces model complexity by employing 1 × 1 convolutions in deeper layers. Experimental results show that the proposed customized architecture significantly improves detection accuracy compared to other state-of-the-art methods, particularly in challenging small-block scenarios, and demonstrates effectiveness in detecting both double and multiple JPEG compression.

关键词：

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：