检索结果-内蒙古大学图书馆

arXiv 2025年

作者： Wang, Lei Li, Xudong Zhang, Han Wang, Jinyi Jiang, Dingkang Xue, Zhidong Wang, Yan School of Software Engineering Huazhong University of Science and Technology China School of Life Science and Technology Huazhong University of Science and Technology China School of Computer Science and Technology Huazhong University of Science and Technology China

At the intersection of the rapidly growing biological data landscape and advancements in Natural Language Processing (NLP), protein language models (PLMs) have emerged as a transformative force in modern research. These models have achieved remarkable progress, highlighting the need for timely and comprehensive overviews. However, much of the existing literature focuses narrowly on specific domains, often missing a broader analysis of PLMs. This study provides a systematic review of PLMs from a macro perspective, covering key historical milestones and current mainstream trends. We focus on the models themselves and their evaluation metrics, exploring aspects such as model architectures, positional encoding, scaling laws, and datasets. In the evaluation section, we discuss benchmarks and downstream applications. To further support ongoing research, we introduce relevant mainstream tools. Lastly, we critically examine the key challenges and limitations in this rapidly evolving field. Copyright © 2025, The Authors. All rights reserved.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Iterative Optimal Attention and Local Model for Single Image Rain Streak Removal

arXiv

引用

arXiv 2025年

作者： Li, Xiangyu Fan, Wanshu Shen, Yue Wang, Cong Wang, Wei Yang, Xin Zhang, Qiang Zhou, Dongsheng National and Local Joint Engineering Laboratory of Computer Aided Design School of Software Engineering Dalian University Dalian China Department of Computing The Hong Kong Polytechnic University Hong Kong School of Cyber Science and Technology Sun Yat-Sen University Shenzhen China School of Computer Science and Technology Dalian University of Technology Dalian China National and Local Joint Engineering Laboratory of Computer Aided Design School of Software Engineering Dalian University School of Computer Science and Technology Dalian University of Technology Dalian China

High-fidelity imaging is crucial for the successful safety supervision and intelligent deployment of vision-based measurement systems (VBMS). It ensures high-quality imaging in VBMS, which is fundamental for reliable visual measurement and analysis. However, imaging quality can be significantly impaired by adverse weather conditions, particularly rain, leading to blurred images and reduced contrast. Such impairments increase the risk of inaccurate evaluations and misinterpretations in VBMS. To address these limitations, we propose an Expectation Maximization Reconstruction Transformer (EMResformer) for single image rain streak removal. The EMResformer retains the key self-attention values for feature aggregation, enhancing local features to produce superior image reconstruction. Specifically, we propose an Expectation Maximization Block seamlessly integrated into the single image rain streak removal network, enhancing its ability to eliminate superfluous information and restore a cleaner background image. Additionally, to further enhance local information for improved detail rendition, we introduce a Local Model Residual Block, which integrates two local model blocks along with a sequence of convolutions and activation functions. This integration synergistically facilitates the extraction of more pertinent features for enhanced single image rain streak removal. Extensive experiments validate that our proposed EMResformer surpasses current state-of-the-art single image rain streak removal methods on both synthetic and real-world datasets, achieving an improved balance between model complexity and single image deraining performance. Furthermore, we evaluate the effectiveness of our method in VBMS scenarios, demonstrating that high-quality imaging significantly improves the accuracy and reliability of VBMS tasks. Codes are at https://***/ghfkahfk/EMResformer. Copyright © 2025, The Authors. All rights reserved.

关键词： Rain

来源：评论

学校读者我要写书评

暂无评论

Enhancing Retrieval Augmented Generation Systems Using AI Models and Graph Databases

Enhancing Retrieval Augmented Generation Systems Using AI Mo...

引用

Emerging Smart Computing and Informatics (ESCI), Conference on

作者： Souvik Mukherjee Soosan Shabnam Sara Hasan D. Ajitha Department of Computer Science and Engineering Vellore Institute of Technology Vellore India Department of Software Systems School of Computer Science and Engineering (SCOPE) Vellore Institute of Technology Vellore Tamil Nadu India

ISBN: (数字)9798331515683

ISBN: (纸本)9798331515690

This paper outlines the process of generating a Neo4j graph database powered by Language Models (LLMs). The primary goal is to extract structured information from unstructured data, including user profiles, paper briefs, and Slack messages, and convert them into Cypher queries. The data is then ingested into Neo4j to build a graph database that captures relationships between users, paper, technologies, and messages. A pipeline was developed to automate the process, ensuring accurate entity and relationship extraction using predefined templates. This approach allows for efficient data representation and supports consultancy in managing large datasets by generating insightful visualizations and querying capabilities.

关键词： Ciphers Accuracy Databases Pipelines Retrieval augmented generation Loading Data visualization Transforms Paper technology Data mining

来源：评论

学校读者我要写书评

暂无评论

Prompt-Driven Continual Graph Learning

arXiv

引用

arXiv 2025年

作者： Wang, Qi Zhou, Tianfei Yuan, Ye Mao, Rui Beijing Institute of Technology Zhuhai China School of Computer Science and Technology Beijing Institute of Technology Beijing China College of Computer Science and Software Engineering Shenzhen University Shenzhen China

Continual Graph Learning (CGL), which aims to accommodate new tasks over evolving graph data without forgetting prior knowledge, is garnering significant research interest. Mainstream solutions adopt the memory replay-based idea, i.e., caching representative data from earlier tasks for retraining the graph model. However, this strategy struggles with scalability issues for constantly evolving graphs and raises concerns regarding data privacy. Inspired by recent advancements in the prompt-based learning paradigm, this paper introduces a novel prompt-driven continual graph learning (PROMPTCGL) framework, which learns a separate prompt for each incoming task and maintains the underlying graph neural network model fixed. In this way, PROMPTCGL naturally avoids catastrophic forgetting of knowledge from previous tasks. More specifically, we propose hierarchical prompting to instruct the model from both feature- and topology-level to fully address the variability of task graphs in dynamic continual learning. Additionally, we develop a personalized prompt generator to generate tailored prompts for each graph node while minimizing the number of prompts needed, leading to constant memory consumption regardless of the graph scale. Extensive experiments on four benchmarks show that PROMPTCGL achieves superior performance against existing CGL approaches while significantly reducing memory consumption. Our code is available at https://***/QiWang98/PromptCGL. © 2025, CC BY.

关键词： Knowledge graph

来源：评论

学校读者我要写书评

暂无评论

Class-Aware Prompting for Federated Few-Shot Class-Incremental Learning

引用

IEEE Transactions on Circuits and Systems for Video technology 2025年

作者： Liang, Fang-Yi Zhan, Yu-Wei Liu, Jiale Zhang, Chong-Yu Chen, Zhen-Duo Luo, Xin Xu, Xin-Shun Shandong University School of Software Jinan250101 China Tsinghua University Department of Computer Science and Technology Beijing100084 China Yunnan Key Laboratory of Software Engineering Kunming650504 China

Few-Shot Class-Incremental Learning (FSCIL) aims to continuously learn new classes from limited samples while preventing catastrophic forgetting. With the increasing distribution of learning data across different clients and privacy concerns, FSCIL faces a more realistic scenario where few learning samples are distributed across different clients, thereby necessitating a Federated Few-Shot Class-Incremental Learning (FedFSCIL) scenario. However, this integration faces challenges from non-IID problem, which affects model generalization and training efficiency. The communication overhead in federated settings also presents a significant challenge. To address these issues, we propose Class-Aware Prompting for Federated Few-Shot Class-Incremental Learning (FedCAP). Our framework leverages pre-trained models enhanced by a class-wise prompt pool, where shared class-wise keys enable clients to utilize global class information during training. This unifies the understanding of base class features across clients and enhances model consistency. We further incorporate a class-level information fusion module to improve class representation and model generalization. Our approach requires very few parameter transmission during model aggregation, ensuring communication efficiency. To our knowledge, this is the first study to explore the scenario of FedFSCIL. Consequently, we designed comprehensive experimental setups and made the code publicly available. © 2025 IEEE. All rights reserved.

关键词： Federated learning

来源：评论

学校读者我要写书评

暂无评论

Human–object interaction detector with unsupervised domain adaptation

引用

Knowledge-Based Systems 2025年 321卷

作者： Cheng, Yamin Duan, Zhekai Huang, Hualong Wang, Zhi School of Computer and Software Engineering Xihua University Chengdu China Department of Computer Science University College London United Kingdom School of Computer Science and Engineering University of Electronic Science and Technology of China Chengdu China

Human–object interaction (HOI) detection constitutes a foundational challenge in fine-grained visual scene understanding. While mainstream HOI detection frameworks rely on supervised training of deep neural networks with extensive labeled datasets, such data-intensive paradigms face practical limitations due to the high annotation and costs in real-world applications. To tackle this problem, it is a natural idea to leverage data from a well-labeled domain (source domain) to improve the performance for the domain of interest (target domain). However, the domain shift problem renders it non-trivial to transfer knowledge between different domains. In this paper, we propose a novel Domain-Adaptive HOI Detection Framework that effectively bridges domain gaps and achieves robust performance on unlabeled target domains by leveraging knowledge from labeled source domains. Specifically, our method aims to align features on both the map-level and sequence-level to learn domain-invariant features, and minimize the interaction-aware feature discrepancy by learning prototypical patterns of action relationship. To this end, we introduce a new adversarial training schema that uses a CutMix strategy to minimize the domain shift of map-level feature representations from the feature backbone. We then delve into feature distribution alignment and pursue sequence-level domain-invariant features from the transformer encoder by designing a token-based feature alignment. In addition, we design a memory bank to minimizes the distance between each inter-class interaction feature with homogeneous relationship, and maximizes the discrepancy between inter-class interaction features with heterogenous relationship. Extensive experiments on multiple benchmarks demonstrate that our proposed framework significantly outperforms prior supervised learning methods and achieves the superior performance. © 2025 Elsevier B.V.

关键词： Deep neural networks

来源：评论

学校读者我要写书评

暂无评论

Attention-guided CenterNet deep learning approach for lung cancer detection

引用

computers in Biology and Medicine 2025年 186卷 109613-109613页

作者： Dawood, Hussain Nawaz, Marriam Ilyas, Muhammad U. Nazir, Tahira Javed, Ali School of Computing Skyline University College Sharjah United Arab Emirates Department of Software Engineering University of Engineering and Technology-Taxila Punjab 47050 Pakistan School of Computer Science University of Birmingham Dubai United Arab Emirates Department of Software Engineering and Computer Science Riphah International University Gulberg Green Campus Islamabad Pakistan

Lung cancer remains a significant health concern worldwide, prompting ongoing research efforts to enhance early detection and diagnosis. Prior studies have identified key challenges in existing approaches, including limitations in feature extraction, interpretability, and computational efficiency. In response, this study introduces a novel deep learning (DL) framework, termed the Improved CenterNet approach, tailored specifically for lung cancer detection. The primary importance of this work lies in its innovative integration of ResNet-34 with an attention mechanism within the CenterNet architecture, addressing critical limitations identified in previous studies. By augmenting the base network with an attention mechanism, our framework offers improved feature extraction capabilities, enabling the model to learn relevant patterns associated with lung cancer amidst complex backgrounds and varying environmental conditions. This enhancement facilitates more accurate and interpretable predictions while reducing computational complexity and inference times. Through extensive experimental evaluations conducted on standard datasets, our proposed approach demonstrates promising results, highlighting its potential to advance the field of lung cancer detection and diagnosis. Specifically, we have acquired the precision, recall, and F1-Score of 99.89 %, 99.82 %, and 99.85 % on the LUNA-16 dataset, and 98.33 %, 98.02 %, and 98.17 % for the Kaggle data sample, respectively which is showing the efficacy of our approach. One limitation of the work is that it cannot effectively locate the samples with intense light variations. Therefore, future research work is focused on overcoming this challenge. © 2024 Elsevier Ltd

关键词： Lung cancer

来源：评论

学校读者我要写书评

暂无评论

Privacy protection and data security in intelligent recommendation systems

引用

Neural Computing and Applications 2025年 1-18页

作者： Zheng, Donghua Alkawaz, Mohammed Hazim Johar, Md Gapar Md School of Information Technology and Engineering Guangzhou College of Commerce Guangdong Guangzhou511363 China School of Graduate Studies Management and Science University Selangor Shah Alam40100 Malaysia Department of Computer Science College of Education for Pure Science University of Mosul Nineveh Mosul41002 Iraq Software Engineering and Digital Innovation Centre Management and Science University Selangor Shah Alam40100 Malaysia

Data security and privacy have emerged as major areas of study for recommendation systems. This is crucial for enhancing system security given the quick advancement of Internet technology and its increased use. However, there are still issues with privacy protection and data security in current intelligent recommendation systems, such as a high probability of privacy leakage, low data security, and poor recommendation performance. To better promote the wider application of intelligent recommendation systems, this article introduces federated learning and differential privacy technology to conduct in-depth research on privacy protection and data security in intelligent recommendation systems. In the article, a federated learning framework is used to construct an intelligent recommendation system model, and the constructed model is trained. Furthermore, a differential privacy mechanism is introduced, which can protect the privacy of samples by analyzing the sensitivity of L1 and L2, gradient descent, gradient pruning, and other methods. Next, the secure multi-party computation (SMC) protocol can be used to achieve model parameter sharing and protect the privacy of participants;subsequently, there is the implementation of federated learning. Finally, this article also evaluates the practical application effects of federated learning and differential privacy technology in privacy protection and data security in intelligent recommendation systems. The study’s findings indicate that the approach suggested in this article has an average data security and privacy leakage risk of 5.35% and 96.64%;federated learning is 16.55% and 85.08%, respectively;differential privacy is 9.65% and 88.47%, respectively;deep learning is 21.08% and 77.92%, respectively;and homomorphic encryption is 10.69% and 82.43%, respectively. The combination of federal learning and differential privacy can effectively protect the privacy and data security of the intelligent recommendation system, reduce

关键词： Sensitive data

来源：评论

学校读者我要写书评

暂无评论

Bluetooth Vulnerabilities and Security

Bluetooth Vulnerabilities and Security

引用

Emerging Smart Computing and Informatics (ESCI), Conference on

作者： Mareeswari V Vijayan R Ashish Shukla Department of Software and Systems Engineering School of Computer Science Engineering and Information Systems (SCORE) Vellore Institute of Technology (VIT) Vellore India Department of Information Technology School of Computer Science Engineering and Information Systems (SCORE) Vellore Institute of Technology (VIT) Vellore India Master of Computer Applications School of Computer Science Engineering and Information Systems (SCORE) Vellore Institute of Technology (VIT) Vellore India

ISBN: (数字)9798331515683

ISBN: (纸本)9798331515690

Bluetooth technology, which facilitates wireless communication between Billions of devices including smartphones, tablets, laptops, and Internet of Thing (IoT) devices, is a cornerstone of modern connectivity. Its importance lies in its ability to enable seamless data exchange and interaction across a wide range of applications, from personal gadgets to complex industrial systems. Despite its widespread adoption, Bluetooth is not immune to critical security vulnerabilities. Issues like Bluetooth Low Energy (BLE) vulnerabilities and denial-of-service (DoS) attacks can overwhelm devices with excessive traffic, while BLE Forced Connection can lead to unauthorized access, and eavesdropping exposes sensitive data being exchanged. We have discussed these vulnerabilities in detail, examining their mechanisms, potential impacts, and various implementation aspects. Additionally, we have added demonstrations to illustrate how these attacks can be executed and the potential consequences. To prevent these threats, we explored essential measures, including regular firmware updates, secure pairing protocols, and careful management of Bluetooth settings. Adopting best practices like disabling Bluetooth when not in use and monitoring connected devices can further enhance security, ensuring the reliable and safe operation of the vast and growing ecosystem of Bluetooth-enabled devices.

关键词： Wireless communication Ethics Protocols Ecosystems Denial-of-service attack Security Internet of Things Bluetooth Low Energy Microprogramming Guidelines

来源：评论

学校读者我要写书评

暂无评论

A novel deep learning approach for investigating liquid fuel injection in combustion system

引用

Discover Artificial Intelligence 2025年第1期5卷 1-14页

作者： Inam, Syed Azeem Khan, Abdullah Ayub Ahmed, Noor Mazhar, Tehseen Shahzad, Tariq Khan, Sunawar Saeed, Mamoon M. Hamam, Habib Department of Artificial Intelligence and Mathematical Science Sindh Madressatul Islam University Sindh Karachi Pakistan Department of Computer Science and Information Technology Shaheed Benazir Bhutto University Lyari Karachi Pakistan Department of Computer Science Shaheed Zulfiqar Ali Bhutto Institute of Science and Technology Karachi Pakistan Department of Computer Science School Education Department Government of Punjab Layyah31200 Pakistan School of Computer Science National College of Business Administration and Economics Lahore54000 Pakistan Department of Computer Engineering COMSATS University Islamabad Sahiwal Campus Sahiwal57000 Pakistan Department of Software Engineering Islamia University of Bahawalpur Bahawalnagar62300 Pakistan Sana’a00967 Yemen Faculty of Engineering Uni de Moncton MonctonNBE1A3E9 Canada Department of Electrical and Electronic Engineering Sciences University of Johannesburg Johannesburg South Africa Av. Grandes Ecoles BP 1989 Libreville Gabon Bridges for Academic Excellence Spectrum Center-ville Tunis Tunisia

The intricacies and instability of introducing cryogenic propellants into the combustion system have piqued the curiosity of scientists studying the procedure. The latest innovation is utilizing data-driven machine learning and deep learning approaches to gain deeper insights into the related difficulties. However, the current work serves as a baseline for future research because relatively few studies have used data-driven methodologies to assess the temperature of liquid fuel injections in combustion systems. The performance of Linear Regression (LR), Random Forest (RF), Extra Trees Regressor (ETR), Polynomial Regression (PR), Support Vector Regressor (SVR), Decision Tree Regressor (DTR), Gradient Boost Regressor (GBR), XGB Regressor (XGBoost), AdaBoost Regressor (ABR), K-Neighbors Regressor (KNR), Long-Short Term Memory (LSTM), Bi-LSTM (Bi-directional Long-Short Term Memory) has all been investigated in this study. The study also suggested a Fully Connected Neural Network (FCNN) to examine its performance and paired it with an Extra Tree Regressor (ETR). The coupled FCNN and Extra Tree Regressor outperform the other algorithms with a Mean Square Error (MSE) of 0.0000005062, Root Mean Square Error (RMSE) of 0.00071148, Mean Absolute Error (MAE) of 0.00020672, and R-squared (R2) value of 0.99998689. Linear Regression, Polynomial Regression, and Support Vector Regressor are found to be the least-performing algorithms. The current work uses machine learning and deep learning methods to make data-driven decisions for liquid fuel injection in the combustion system. © The Author(s) 2025.

关键词： Deep learning

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：