检索结果-内蒙古大学图书馆

arXiv 2025年

作者： Wang, Zhenghua Ding, Yiran Lv, Changze Xu, Zhibo Li, Tianlong Shi, Tianyuan Zheng, Xiaoqing Huang, Xuanjing Fudan University China Westlake University China Shanghai Key Laboratory of Intelligent Information Processing China

Although large language models (LLMs) have achieved significant progress in handling long-context inputs, they still suffer from the "lost-in-the-middle" problem, where crucial information in the middle of the context is often underrepresented or lost. Our extensive experiments reveal that this issue may arise from the rapid long-term decay in Rotary Position Embedding (RoPE). To address this problem, we propose a layer-specific positional encoding scaling method that assigns distinct scaling factors to each layer, slowing down the decay rate caused by RoPE to make the model pay more attention to the middle context. A specially designed genetic algorithm is employed to efficiently select the optimal scaling factors for each layer by incorporating Bézier curves to reduce the search space. Through comprehensive experimentation, we demonstrate that our method significantly alleviates the "lost-in-the-middle" problem. Our approach results in an average accuracy improvement of up to 20% on the key-Value Retrieval dataset. Furthermore, we show that layer-specific interpolation, as opposed to uniform interpolation across all layers, enhances the model’s extrapolation capabilities when combined with PI and Dynamic-NTK positional encoding schemes. Copyright © 2025, The Authors. All rights reserved.

关键词： Rope

来源：评论

学校读者我要写书评

暂无评论

FGCM: Modality-behavior Fusion Model Integrated with Graph Contrastive Learning for Multimodal Recommendation

引用

IEEE Multimedia 2025年

作者： Mo, Feng Xiao, Lin Song, Qiya Gao, Xieping Song, Wenzhuo Wang, Shoujin Hunan Normal University Hunan Provincial Key Laboratory of Intelligent Computing and Language Information Processing Changsha410081 China Northeast Normal University Changchun China University of Technology Sydney Sydney Australia

Multimodal recommender systems (MRSs) aim to integrate information from multiple modalities, for better capturing users' preferences. However, existing MRSs usually face the challenge of data sparsity, especially for the limited user-item interactions. It is not uncommon that a user interacts with only a few items out of millions. More importantly, the interaction information is easy to be overwhelmed, leading to poor recommendation performance. To tackle these significant gaps for further improving the performance of MRSs, we propose a Modality-behavior Fusion Model Integrated with Graph Contrastive Learning for Multimodal Recommendation (FGCM). First, through the introduction of random noise into graph contrastive learning, FGCM adeptly mitigates data sparsity. Then, a modality-behavior fuser is proposed to effectively fuse interaction data and multimodal side information data by taking cooperative signals and multimodal signals as the guidance. Extensive experiments on three datasets demonstrate that our proposed approach outperforms several state-of-the-art methods. © 2025 IEEE.

关键词： Contrastive Learning

来源：评论

学校读者我要写书评

暂无评论

Recommendation-Aware Collaborative Edge Caching Strategy in the Internet of Vehicles 18th

Recommendation-Aware Collaborative Edge Caching Strategy in...

引用

18th International Conference on Wireless Artificial intelligent computing Systems and Applications, WASA 2024

作者： Ou, Pingjie Chen, Ningjiang Huang, Zizhan School of Computer and Electronic Information Guangxi University Guangxi Nanning530004 China Key Laboratory of Parallel Distributed and Intelligent Computing Education Department of Guangxi Zhuang Autonomous Region Guangxi University Guangxi Nanning530004 China

ISBN: (纸本)9783031714696

Due to the limited resources of edge networks, the heterogeneity of user content requests, high-cost caching from direct resource hits, and redundancy in resource retention time hinder system performance. Traditional caching methods based on learning and popularity tend to encounter data skew when handling vehicle requests, leading to untimely cache data updates. To address these issues, this paper proposes an efficient edge cooperative caching strategy with recommendation awareness. Firstly, a time series prediction model based on a multi-dense multilayer perceptron is designed. This model segments historical request information into sub-sequence level slivers to enhance positional correlations between information, predicting user requests through historical time series and covariates. Secondly, to optimize cache utilization, comprehensively considers user mobility and contribution to enable recommendation awareness. Additionally, an optimized lifecycle is introduced to design a cache replacement strategy. Simulation experiments demonstrate that, compared to caching schemes based on content popularity and deep reinforcement learning, the proposed method achieves lower caching costs and higher cache hit rates. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

关键词： Deep reinforcement learning

来源：评论

学校读者我要写书评

暂无评论

Active Security Control for Networked Jumping Systems Under Asynchronous Dual-channel DoS Attacks: An HMM-based Approach

引用

IEEE Transactions on Automation Science and Engineering 2025年 22卷 14976-14986页

作者： Cheng, Peng Ye, Hu Wu, Di Zhang, Weidong Anhui University Key Laboratory of Intelligent Computing and Signal Processing of Ministry of Education School of Electrical Engineering and Automation Hefei230601 China Shanghai Jiao Tong University Department of Automation Shanghai200240 China Hainan University School of Electronic Science and Technology Hainan Haikou570228 China

This paper aims to address the design problem of asynchronous controllers for networked jumping systems (NJSs), in which two communication channels are vulnerable to asynchronous denial-of-service (DoS) attacks. To this end, a hidden Markov model (HMM)-based active security control strategy is proposed to counter asynchronous dual-channel DoS attacks. Firstly, two maximum consecutive numbers are introduced to describe DoS attacks randomly initiated by adversaries. Subsequently, a mode-dependent predictor is designed to generate predictive states, which can be employed in the switched controller to stabilize the NJSs. Furthermore, sufficient conditions are derived by constructing Lyapunov functions to ensure the asymptotic mean-square stability of the NJSs under asynchronous dual-channel DoS attacks. Finally, a simulation using a space robot manipulator model is presented to validate the effectiveness and practicality of the proposed active security control strategy. © 2004-2012 IEEE.

关键词： Active safety systems

来源：评论

学校读者我要写书评

暂无评论

Application of the CNN-GRU-MHA Model in Partial Discharge Recognition 24

Application of the CNN-GRU-MHA Model in Partial Discharge Re...

引用

2nd International Conference on Advances in Artificial Intelligence and Applications, AAIA 2024

作者： Bi, Jinjie Chen, Guofu Li, Tinghui Huang, Shoulin Shen, Biquan Yang, Rixing Guangxi Key Laboratory of Brain-inspired Computing and Intelligent Chips School of Electronic and Information Engineering Guangxi Normal University Guilin China Education Department of Guangxi Zhuang Autonomous Region Guangxi Normal University Guilin China

ISBN: (纸本)9798400712883

Partial discharge (PD) is one of the most important factors affecting the stable operation of high-voltage equipment and poses a great threat to the reliability and safety of power systems. Although some deep learning-based methods, such as Convolutional Neural Networks (CNN), have been proposed for PD detection, there are still some shortcomings in terms of global dependencies. In this paper, it is proposed to develop a model of Convolutional Neural Networks-Gated Recurrent Unit-Multi-Head Attention (CNN-GRU-MHA). Wavelet transformation is adopted for noise suppression. reduction, together with statistical feature downsampling for reducing data dimensionality. In the model, CNN extract localized temporal features, GRU capture sequential temporal patterns, and MHA focuses on modeling global dependencies while highlighting critical features. Experimental evaluations reveal that the model reaches a Matthews correlation coefficient of 74.78% and a recall rate of 74.61%, performing better than some conventional methods in terms of both accuracy and robustness. This model securely and effectively provided a method of enhancing power system safety and stability. © 2024 Copyright held by the owner/author(s).

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

DiffTimb: Diffusion Models for Many-to-Many Timbre Transfer 10th

DiffTimb: Diffusion Models for Many-to-Many Timbre Transfer

引用

10th Conference on Sound and Music Technology, CSMT 2023

作者： Gan, Xu Wu, Yifei Liu, Xinlu Duan, Ruilei Li, Wei School of Computer Science and Technology Fudan University Shanghai200438 China Department of Music Engineering ZheJiang Conservatory of Music Hangzhou310024 China Shanghai Key Laboratory of Intelligent Information Processing Fudan University Shanghai200433 China

ISBN: (纸本)9789819779611

The goal of the timbre transfer task is to apply the auditory characteristics of one instrument to another instrument. Past timbre transfer methods faced numerous challenges due to the limitations of the selected generation model. These challenges include difficulties in training, ambiguous results, poor diversity, and slow generation speed. To address these issues, we propose a diffusion model for timbre transfer, DiffTimb, which can be combined with vocoder to achieve single-note timbre transfer. Unlike traditional models that require training a specific model for each pair of timbre transfers, DiffTimb enables the conversion of various musical instrument timbres received in a single model, namely, many-to-many timbre transfer. We compare two Unet-based denoising frameworks and propose two feature fusion methods to fuse timbre and pitch features. We conduct unconditional and conditional generation pre-experiments to verify the generation ability of DiffTimb. To improve the generation effect, we clean the Nsynth dataset and get two subsets for our task. Experimental results show that DiffTimb has outperformed previous methods in both subjective and objective evaluations. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025.

关键词： Audio acoustics

来源：评论

学校读者我要写书评

暂无评论

A New Dual-Population Evolutionary Algorithm Leveraging Objective-Constraint Relationships for Constrained Multi-Objective Optimization

SSRN

引用

SSRN 2025年

作者： Ye, Jialu Tan, Chaogui Xia, Yizhang Hou, Zhanglu Liu, Yuan Zou, Juan Hunan Engineering Research Center of intelligent System Optimization and Security Key Laboratory of Intelligent Computing and Information Processing Ministry of Education Hunan National Applied Mathematics Center China Xiangtan University China

Population co-evolution strategies are widely used to handle constrained multi-objective optimization problems (CMOPs). However, existing coevolutionary algorithms oversimplify population collaboration and are rigid in evolving the auxiliary population, ignoring the importance of maintaining population diversity, especially for CMOPs with irregular feasible regions. To tackle this issue, we propose a new dual-population evolutionary algorithm, denoted as AACMO, which attempts to mine the relational features between the unconstrained Pareto front (UPF) and the constrained Pareto front (CPF) to better adapt to CMOPs with various constraints. Specifically, the method is categorized into two phases: a learning phase and an evolving phase. First, the main population (MP) aims to quickly learn the relationships between the UPF and the CPF by leveraging the advantages of different operators, namely, genetic algorithm operators and differential evolution operators, during the learning phase. Secondly, two auxiliary strategies are designed to collaborate with the MP during the evolving phase. These strategies are capable of employing a more effective co-evolution method to assist the MP in maintaining diversity, based on the degree of alignment between the learned UPF and CPF at the end of the learning phase. The effectiveness of AACMO is confirmed through comparative analysis with eight state-of-the-art algorithms across four CMOP benchmark suites. © 2025, The Authors. All rights reserved.

关键词： Optimization algorithms

来源：评论

学校读者我要写书评

暂无评论

Surface-PMSM Sensorless Control Strategy Based on Reduced-Order Linear Kalman Filter Cooperating with Prediction Error Rolling Compensation

引用

IEEE Transactions on Power Electronics 2025年

作者： Xu, Sang Shen, Anwen Tang, Qipeng Wang, Miaomiao Luo, Pan Luo, Xin Xu, Jinbang National Key Laboratory of Science and Technology on Multispectral Information Processing Key Laboratory of Image Processing and Intelligent Control School of Artificial Intelligence and Automation Wuhan430074 China Science and Technology on Reactor System Design Technology Laboratory Chengdu China

Kalman filter (KF) is increasingly attracted for sensorless control of surface permanent magnet synchronous motors (SPMSMs) due to its strong robustness against measurement and system noise. However, conventional method suffers from poor position and speed estimation accuracy under dynamic conditions, and high computational cost. In order to solve these problems, an enhanced sensorless control strategy based on reduced-order linear Kalman filter (RLKF) cooperating with prediction error rolling compensation is introduced in this paper. Differing from conventional method, a reduced-order linear state equation encompassing rotor position, speed and torque disturbance is constructed. Then, by obtaining the prediction error of system output using the motor current equation in the predictive reference frame, the optimal estimation of rotor position and speed in KF algorithm is achieved based on the reduced-order model. This not only simplifies the calculation process, but also enhances the position and speed estimation accuracy under dynamic conditions. Meanwhile, a prediction error rolling compensation algorithm is developed to minimize the estimation deviations caused by motor parameters variations. Additionally, the system's dynamic performance is further improved with an adopted speed controller based on torque disturbance feedforward and acceleration feedback. Finally, experimental results verify the effectiveness and feasibility of the proposed strategy.1 © 1986-2012 IEEE.

关键词： Permanent magnets

来源：评论

学校读者我要写书评

暂无评论

Multi-scale Traffic Camera Image Detection Network Based on Improved YOLOv8 21st

Multi-scale Traffic Camera Image Detection Network Based on...

引用

21st Pacific Rim International Conference on Artificial Intelligence, PRICAI 2024

作者： Peng, Zhihao Qi, Xinyuan Wu, Sheng Shang, Jianga Yang, Linquan China University of Geosciences Wuhan430074 China Engineering Research Center of Natural Resource Information Management and Digital Twin Engineering Software Ministry of Education Wuhan430074 China Hubei Key Laboratory of Intelligent Geo-Information Processing China University of Geosciences Wuhan430074 China

ISBN: (纸本)9789819601240

With the extensive deployment of traffic cameras across cities, utilizing these devices for vehicle detection has become essential for efficient traffic management. However, variations in vehicle sizes and distances between vehicles and cameras lead to inconsistent scaling of vehicles in the images, which presents a challenge for feature extraction. Furthermore, conventional object detection algorithms fail to fully utilize the rich feature information across multiple layers. To solve these issues, we develop a novel and advanced multi-scale traffic YOLO (MT-YOLO), based on the YOLOv8 model. Firstly, we propose the spatial multi-scale (SMS) attention mechanism to improve the extraction capability of spatial multi-scale vehicle features and aggregate these features for dynamic spatial modulation. Secondly, we develop the multi-scale dilation transposition CARAFE (MDT-CARAFE) upsampling method, which utilizes dilated convolutions to enhance the receptive field at different scales during the upsampling process. Lastly, we augment the network with a detection head and replace the original detection head with the adaptively spatial feature fusion(ASFF) detection head, which dynamically selects and fuses multi-layer feature maps, enabling the network to better capture multi-scale information. Our experiments demonstrate that the MT-YOLO model greatly improves results on the traffic camera detection dataset., achieving an mAP@0.5 of 80.9%, with a precision of 92% and recall of 93%. These results surpass the performance of leading detection algorithms. In addition, the model takes only 12.5 milliseconds to complete an inference, which makes it highly deployable and valuable for real-world applications. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025.

关键词： Image enhancement

来源：评论

学校读者我要写书评

暂无评论

Pencil beam based on the computational holography method 2

Pencil beam based on the computational holography method

引用

2nd International Academic Conference on Optics and Photonics, IACOP 2024

作者： Luo, Tianlei Lv, Anzhi Liu, Minghao Dong, Jiaqing Liu, Guolin Li, Zilong Song, Xianlin School of Information Engineering Nanchang University Nanchang330031 China School of Advanced Manufacture Nanchang University Nanchang330031 China Jiangxi Provincial Key Laboratory of Advanced Signal Processing and Intelligent Communications Nanchang University Nanchang330031 China

ISBN: (数字)9781510688827

ISBN: (纸本)9781510688810

This paper presents a novel pencil beam generation method based on computational holography. The method generates intensity maps through computer simulation, obtains a computed hologram using a computational holographic method, and simulates near-field diffraction using computer simulation to simulate the role of a spatial light modulator (SLM) in a real holographic system. The theoretical effects of the computed hologram are simulated using computer simulation, and finally, a pencil-shaped beam with a focus position that can be flexibly adjusted at all depths is generated. Through experimental simulations, the feasibility of adjusting the focus position of the beam within a certain depth range is verified, showing its potential application value in fluorescence excitation and biomedical fields. © 2025 SPIE.

关键词： Holograms

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：