检索结果-内蒙古大学图书馆

Improved 3-D LSTM: A Video Prediction Approach to Long Sequence Load Forecasting

IEEE TRANSACTIONS ON SMART GRID 2025年第2期16卷 1885-1896页

作者： Xiao, Jiang-Wen Cui, Xue-Ying Liu, Xiao-Kang Fang, Hongliang Li, Peng-Cheng Huazhong Univ Sci & Technol Sch Artificial Intelligence & Automat Key Lab Image Proc & Intelligent Control Wuhan 430074 Peoples R China

Power load forecasting is the foundation of maintaining power grid stability, and can assist in decision-making to reduce operating costs. Fine-grained long sequence load forecasting contributes to formulating plans for power purchase, electricity consumption, energy storage, etc. Long sequence load forecasting requires models to effectively store memory and to accurately capture the long-term complex mapping between output and input. Therefore, this paper converts load sequences into three-dimensional (3D) continuous video frames and presents a model based on long short-term memory (LSTM) named the Improved 3D LSTM (I3D-LSTM) for predicting video frames. It contains two 3D LSTM units: For highly periodic load data, a Long-memory 3D LSTM unit is proposed, which has stronger long-term memory and removes short-term memory;On weakly periodic datasets, a Simplified 3D LSTM unit without the scoring parts exhibits excellent performance. I3D-LSTM also contains a 3D recurrent neural network architecture with residual. Dropblock and batch normalization are integrated into the I3D-LSTM, which are analyzed as excellent solutions for overfitting in 3D LSTM. Comprehensive tests are conducted on different sequence lengths in multiple real-world datasets. Comparison results indicate that I3D-LSTM outperforms various advanced models.

关键词： Long short term memory Three-dimensional displays Load modeling Forecasting Load forecasting Convolution Time series analysis Long sequence time series forecasting load forecasting long short-term memory deep learning

来源：评论

学校读者我要写书评

暂无评论

Differentially Private Linearized ADMM Algorithm for Decentralized Nonconvex Optimization

引用

IEEE TRANSACTIONS ON INFORMATION FORENSICS AND SECURITY 2025年 20卷 3316-3329页

作者： Yue, Xiao-Yu Xiao, Jiang-Wen Liu, Xiao-Kang Wang, Yan-Wu Huazhong Univ Sci & Technol Sch Artificial Intelligence & Automat Key Lab Image Proc & Intelligent Control Wuhan 430074 Peoples R China

Privacy preservation is a challenging problem in decentralized nonconvex optimization containing sensitive data. Prior approaches to decentralized nonconvex optimization are either not strong enough to protect privacy or exhibit low utility under a high privacy guarantee. To address these issues, we propose a differentially private linearized alternating direction method of multipliers (DP-LADMM), which achieves fast convergence property for nonconvex objective functions while achieving saddle/maximum avoidance under differential privacy guarantee. We also apply the Analytic Gaussian Mechanism to track the cumulative privacy loss and provide a tight global differential privacy guarantee for DP-LADMM. The theoretical analysis offers an explicit convergence rate for our algorithm. To the best of our knowledge, this is the first paper to provide explicit convergence for decentralized nonconvex optimization with differential privacy and saddle/maximum avoidance. Numerical simulations and comparison studies on decentralized estimation confirm the superiority of the algorithm and the effectiveness of global privacy preservation.

关键词： Optimization Privacy Differential privacy Convergence Linear programming Vectors Machine learning Training Protection Perturbation methods Decentralized nonconvex optimization linearized alternating direction method of multipliers privacy preservation saddle avoidance

来源：评论

学校读者我要写书评

暂无评论

Spatial cascaded clustering and weighted memory for unsupervised person re-identification

引用

image AND VISION COMPUTING 2025年 156卷

作者： Hong, Jiahao Zuo, Jialong Han, Chuchu Zheng, Ruochen Tian, Ming Gao, Changxin Sang, Nong Huazhong Univ Sci & Technol Sch Artificial Intelligence & Automat Key Lab Image Proc & Intelligent Control Wuhan Hubei Peoples R China

Recent advancements in unsupervised person re-identification (re-ID) methods have demonstrated high performance by leveraging fine-grained local context, often referred to as part-based methods. However, many existing part-based methods rely on horizontal division to obtain local contexts, leading to misalignment issues caused by various human poses. Moreover, misalignment of semantic information within part features hampers the effectiveness of metric learning, thereby limiting the potential of part-based methods. These challenges result in under-utilization of part features in existing approaches. To address these issues, we introduce the Spatial Cascaded Clustering and Weighted Memory (SCWM) method. SCWM aims to parse and align more accurate local contexts for different human body parts while allowing the memory module to balance hard example mining and noise suppression. Specifically, we first analyze the issues of foreground omissions and spatial confusions in previous methods. We then propose foreground and space corrections to enhance the completeness and reasonableness of human parsing results. Next, we introduce a weighted memory and utilize two weighting strategies. These strategies address hard sample mining for global features and enhance noise resistance for part features, enabling better utilization of both global and part features. Extensive experiments conducted on Market-1501, DukeMTMC-reID and MSMT17 datasets validate the effectiveness of the proposed method over numerous state-of-the-art methods.

关键词： Person re-identification Fine-grained Unsupervised learning Contrastive learning

来源：评论

学校读者我要写书评

暂无评论

FADE: A Task-Agnostic Upsampling Operator for Encoder-Decoder Architectures

引用

INTERNATIONAL JOURNAL OF COMPUTER VISION 2025年第1期133卷 151-172页

作者： Lu, Hao Liu, Wenze Fu, Hongtao Cao, Zhiguo Huazhong Univ Sci & Technol Sch Artificial Intelligence & Automat Key Lab Image Proc & Intelligent Control Minist Educ Wuhan 430074 Peoples R China

The goal of this work is to develop a task-agnostic feature upsampling operator for dense prediction where the operator is required to facilitate not only region-sensitive tasks like semantic segmentation but also detail-sensitive tasks such as image matting. Prior upsampling operators often can work well in either type of the tasks, but not both. We argue that task-agnostic upsampling should dynamically trade off between semantic preservation and detail delineation, instead of having a bias between the two properties. In this paper, we present FADE, a novel, plug-and-play, lightweight, and task-agnostic upsampling operator by fusing the assets of decoder and encoder features at three levels: (i) considering both the encoder and decoder feature in upsampling kernel generation;(ii) controlling the per-point contribution of the encoder/decoder feature in upsampling kernels with an efficient semi-shift convolutional operator;and (iii) enabling the selective pass of encoder features with a decoder-dependent gating mechanism for compensating details. To improve the practicality of FADE, we additionally study parameter- and memory-efficient implementations of semi-shift convolution. We analyze the upsampling behavior of FADE on toy data and show through large-scale experiments that FADE is task-agnostic with consistent performance improvement on a number of dense prediction tasks with little extra cost. For the first time, we demonstrate robust feature upsampling on both region- and detail-sensitive tasks successfully. Code is made available at: https://***/poppinace/fade

关键词： Feature upsampling Dense prediction Semantic segmentation image matting Object detection Instance segmentation Depth estimation

来源：评论

学校读者我要写书评

暂无评论

Distributed robust iterative learning control for AC microgrids with external disturbances

引用

SCIENCE CHINA-INFORMATION SCIENCES 2025年第10期68卷 1-21页

作者： Su, Hongyu Lai, Jingang Huazhong Univ Sci & Technol Sch Artificial Intelligence & Automat Key Lab Image Proc & Intelligent Control Minist Educ Wuhan 430074 Peoples R China

This study proposes a distributed secondary control scheme based on distributed robust iterative learning control (DRILC) for islanded alternating current (AC) microgrids subjected to external disturbances. By employing the decoupled sliding mode consensus approach, voltage regulation, frequency restoration, and accurate active power sharing can be achieved within a finite time on the proposed novel integral terminal sliding mode (ITSM) manifold. Furthermore, the appropriate iterative update law in the ITSM-based controller is utilized to learn and eliminate external disturbances as effectively as possible. In the proposed control scheme, the iterative learning control and sliding mode control are designed to function in a complementary manner, which enhances the performance of the secondary control scheme for multi-objective regulations. The stability criteria and robustness to external disturbances of the closed-loop microgrid system in the iteration and time domains are also rigorously derived with the help of the Lyapunov direct method. Finally, the effectiveness of the proposed secondary control scheme is validated by case studies of an islanded AC microgrid test system in the MATlab/SimPowerSystems software environment.

关键词： distributed secondary control islanded microgrid iterative learning control sliding mode control external disturbances

来源：评论

学校读者我要写书评

暂无评论

Distributed Reduced-Order Observer-Based Consensus of Multiple Euler-Lagrange Systems Over Switching Networks

引用

IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND ENGINEERING 2025年 22卷 53-61页

作者： Long, Mingkang Su, Housheng Huazhong Univ Sci & Technol Sch Artificial Intelligence & Automat Key Lab Image Proc & Intelligent Control Educ Minist China Wuhan 430074 Peoples R China

In this paper, we investigate leader-follower consensus (LFC) of multiple Euler-Lagrange systems (MELSs) under the case that followers can only measure partial output information of leader system, and the communication network among followers is switching and jointly connected. Firstly, in order to accurately estimate complete state information of the leader system in real time, we propose a distributed reduced-order observer, which is valid over a jointly connected switching network. Then, by applying this distributed reduced-order observer, we further design an observer-based controller for multiple Euler-Lagrange systems (MELSs) to reach leader-follower consensus (LFC). Distinct with the existing controllers for LFC of MELSs, this controller is robust for bounded external disturbances, and independent of the structure and features of Euler-Lagrange (EL) system. At last, a simulation example is given to confirm the effectiveness of the proposed distributed reduced-order observer and robust controller. Note to Practitioners-Many practical engineering applications (such as spacecrafts, mobile robot manipulators) can be modeled as multiple uncertain Euler-Lagrange systems, where each agent suffers external input disturbances, and only a portion of followers can measure partial components of leader's output. However, most relevant works assumed that the dynamic information of each agent is completely known, agents do not suffer external input disturbances, and all follower agents can measure full-dimensional output of leader. How to achieve LFC for such MELSs has become a main focus of control researches. Thus, this paper proposes an adaptive distributed reduced-order observer-based consensus algorithm for such MELSs, whose communication network is switching and jointly connected. Finally, the effectiveness of the proposed algorithm is illustrated by a simulation example.

关键词： Leader-follower consensus reduced-order observer multiple Euler-Lagrange systems model-independent switching networks

来源：评论

学校读者我要写书评

暂无评论

Geometry-sensitive semantic modeling in visual and visual-language domains for image captioning

引用

ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE 2025年 147卷

作者： Zhu, Wencai Jiang, Zetao He, Yuting Guilin Univ Elect Technol Guangxi Key Lab Image & Graph Intelligent Proc Guilin 541004 Peoples R China

Transformer-based models with grid features as visual representations perform well in image captioning. However, the division and flattening operations increase the difficulty of capturing objects and their relationships via pure semantic modeling. Furthermore, the natural language generated by the current Transformer model still suffers from semantic overconcentration. In this paper, we aim to improve the attention modules in two ways to solve the above issues. We first propose a Geometry-Sensitive Self-Attention (GSSA) module, subdivide geometric signals in the visual domain into relative position and distance, and assist the semantic modeling process according to their unique characteristics. It compensates for the lack of objects and their relationships in the grid features. Then, we propose a Geometry-Sensitive Cross-Attention (GSCA) module, which perceives the source neighboring relationships between images and text in the visual-language domain from a geometric perspective and uses these relationships to adjust the semantic correspondences between the two dynamically. It spreads overly focused attention to surrounding grids to improve understanding of full image content during captioning. To prove our designs, we apply GSSA and GSCA to a standard Transformer to construct a novel Geometry-Sensitive Transformer Network (GSTNet), which conducts geometry-sensitive semantic modeling in visual and visual-language domains. Extensive experiments are conducted to verify the effectiveness of our proposal. The results show that our GSTNet achieves superior performance compared to many state-of-theart image captioning models on the Microsoft Common Objects in Context (MSCOCO) dataset. Besides, the generalization of GSTNet is also verified on the Flickr30k dataset.

关键词： image captioning Grid features Transformer mechanism Geometry-sensitive semantic modeling

来源：评论

学校读者我要写书评

暂无评论

DMFormer: Dense Memory Linformer for image Captioning

引用

ELECTRONICS 2025年第9期14卷 1716-1716页

作者： He, Yuting Jiang, Zetao Guilin Univ Elect Technol Guangxi Key Lab Image & Graph Intelligent Proc Guilin 541004 Peoples R China

image captioning is a cross-task of computer vision and natural language processing, aiming to describe image content in natural language. Existing methods still have deficiencies in modeling the spatial location and semantic correlation between image regions. Furthermore, these methods often exhibit insufficient interaction between image features and text features. To address these issues, we propose a Linformer-based image captioning method, the Dense Memory Linformer for image Captioning (DMFormer), which has lower time and space complexity than the traditional Transformer architecture. The DMFormer contains two core modules: the Relation Memory Augmented Encoder (RMAE) and the Dense Memory Augmented Decoder (DMAD). In the RMAE, we propose Relation Memory Augmented Attention (RMAA), which combines explicit spatial perception and implicit spatial perception. It explicitly uses geometric information to model the geometric correlation between image regions and implicitly constructs memory unit matrices to learn the contextual information of image region features. In the DMAD, we introduce Dense Memory Augmented Cross Attention (DMACA). This module fully utilizes the low-level and high-level features generated by the RMAE through dense connections, and constructs memory units to store prior knowledge of image and text. It learns the cross-modal associations between visual and linguistic features through an adaptive gating mechanism. Experimental results on the MS-COCO dataset show that the descriptions generated by the DMFormer are richer and more accurate, with significant improvements in various evaluation metrics compared to mainstream methods.

关键词： image captioning Linformer relation memory augmented attention dense memory augmented cross attention

来源：评论

学校读者我要写书评

暂无评论

Reduced-order interval observer-based coordination control for discrete-time multi-agent systems

引用

AUTOMATICA 2025年 174卷

作者： Su, Housheng Luo, Miaohong Zeng, Zhigang Huazhong Univ Sci & Technol Sch Artificial Intelligence & Automat Image Proc & Intelligent Control Key Lab Educ Minist China Luoyu Rd 1037 Wuhan 430074 Peoples R China

In this paper, the coordination control problem of discrete-time multi-agent systems (MASs) with uncertainties is studied by the output feedback technique, where the uncertainties are unknown disturbances and initial states. Firstly, a reduced-order neighborhood framer is constructed by using the boundary information of uncertainties. Secondly, a control protocol that depends on the absolute information of the agent framer is proposed by solving a modified algebraic Riccati equation. The results demonstrate that the control protocol can render a reduced-order neighborhood framer as a reduced-order neighborhood interval observer, which can not only realize the interval-valued estimation on the sum of the relative states between each agent and its neighbors in real time, but also realize the cooperative behavior of MASs under some sufficient conditions involving network synchronization and the instability degree of the agent. In addition, direct and indirect methods are proposed to eliminate the nonnegative constraint. Finally, the theoretical results are verified by two numerical simulations. (c) 2025 Elsevier Ltd. All rights are reserved, including those for text and data mining, AI training, and similar technologies.

关键词： Coordination control Discrete-time multi-agent systems Uncertainty Reduced-order interval observer

来源：评论

学校读者我要写书评

暂无评论

Consensus of nonlinear multiagent systems on time scales via event-triggered sliding mode control

引用

JOURNAL OF THE FRANKLIN INSTITUTE 2025年第8期362卷

作者： Yang, Chen Tang, Wei Sheng, Yin Zeng, Zhigang Huazhong Univ Sci & Technol Sch Artificial Intelligence & Automat Wuhan 430074 Peoples R China Key Lab Image Proc & Intelligent Control Educ Mini Wuhan 430074 Peoples R China

In this paper, the tracking consensus of timescale-type nonlinear multiagent systems is investigated. First, the finite-time reachability of the designed integral sliding manifold is achieved through a discontinuous controller. Second, combining the proposed dynamic event-triggered strategy, sufficient condition for practical tracking consensus is obtained by constructing non-negative functions and applying inequality techniques. In addition, in combination with time scale calculus theory, Zeno behavior is excluded. The framework for sliding mode control (SMC) of timescale-type multiagent systems is constructed to generalize previous continuous-time SMC and discrete-time SMC results. Ultimately, the feasibility of the obtained theoretical results is validated through two numerical simulations.

关键词： Multiagent systems Sliding mode control Event-triggered control Tracking consensus Time scale

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：