检索结果-内蒙古大学图书馆

Mdsd: a multi-dimensional scaling-based defensive mechanism against backdoor attacks on federated learning

Cluster Computing 2025年第5期28卷 1-22页

作者： Chen, Qiuxian Tao, Yizheng Institute of Computer Application China Academy of Engineering Physics Mianyang China

Federated Learning (FL) is an innovative machine learning paradigm that enables collaborative model training among multiple participants without sharing raw data. Specifically, each participant trains a local model on its dataset and only needs to transmit the model updates or gradient information to build a global model. Although FL helps preserve data privacy, it still faces security concerns, particularly the threat posed by backdoor attacks. Backdoor attacks aim to insert hidden functionalities into the model, causing it to produce predetermined, incorrect outputs when triggered by specific inputs. Current research on backdoor attacks and defenses in FL settings primarily focuses on the image domain. With the growing application of FL in natural language processing, especially large language models, it is necessary to extend the research to these areas. Besides, current defense measures against backdoor attacks have several limitations: detection methods can become inefficient and less effective;mitigation techniques may have limited efficacy or require careful tuning;robust training may add computational overhead and resource requirements. These limitations are more pronounced in high-dimensional models. To counteract backdoor attacks and overcome such limitations, we propose MDSD, a multi-dimensional scaling-based defensive mechanism against backdoor attacks in federated learning. MDSD consists of three components: Clipping with Norm Threshold, Anomaly Detection, and Impact Control. Clipping with Norm Threshold limits the impact from potential malicious models by restricting each local model update under a specific threshold. Anomaly Detection employs the multi-dimensional scaling (MDS) technique to transform high-dimensional models into a lower-dimensional space while preserving the relative distances between data points. In the lower-dimensional space, the local model with the shortest distance to the global model is considered suspicious. Instead of directl

关键词： Federated learning

来源：评论

学校读者我要写书评

暂无评论

Exploring Potential Barrier Estimation Mechanism Based on Quantum Dynamics Framework

引用

Chinese Journal of Electronics 2025年第1期34卷 350-364页

作者： Quan Tang Peng Wang Chengdu Institution of Computer Application Chinese Academy of Science University of Chinese Academy of Sciences School of Computer Science and Engineering Southwest Minzu University

Due to the probability characteristics of quantum mechanism, the combination of quantum mechanism and intelligent algorithm has received wide attention. Quantum dynamics theory uses the Schr?dinger equation as a quantum dynamics equation. Through three approximation of the objective function, quantum dynamics framework(QDF) is obtained which describes basic iterative operations of optimization algorithms. Based on QDF, this paper proposes a potential barrier estimation(PBE) method which originates from quantum mechanism. With the proposed method, the particle can accept inferior solutions during the sampling process according to a probability which is subject to the quantum tunneling effect, to improve the global search capacity of optimization *** effectiveness of the proposed method in the ability of escaping local minima was thoroughly investigated through double well function(DWF), and experiments on two benchmark functions sets show that this method significantly improves the optimization performance of high dimensional complex functions. The PBE method is quantized and easily transplanted to other algorithms to achieve high performance in the future.

关键词： Potential energy Electric potential Shape Heuristic algorithms Estimation Tunneling Linear programming Approximation algorithms Iterative algorithms Optimization

来源：评论

学校读者我要写书评

暂无评论

Deep learning-based open API recommendation for Mashup development

引用

Science China(Information Sciences) 2023年第7期66卷 94-111页

作者： Ye WANG Junwu CHEN Qiao HUANG Xin XIA Bo JIANG School of Computer and Information Engineering Zhejiang Gongshang University Software Engineering Application Technology Lab

Mashup developers often need to find open application programming interfaces(APIs) for their composition application development. Although most enterprises and service organizations have encapsulated their businesses or resources online as open APIs, finding the right high-quality open APIs is not an easy task from a library with several open APIs. To solve this problem, this paper proposes a deep learning-based open API recommendation(DLOAR) approach. First, the hierarchical density-based spatial clustering of applications with a noise topic model is constructed to build topic models for Mashup clusters. Second,developers' requirement keywords are extracted by the Text Rank algorithm, and the language model is built. Third, a neural network-based three-level similarity calculation is performed to find the most relevant open APIs. Finally, we complement the relevant information of open APIs in the recommended list to help developers make better choices. We evaluate the DLOAR approach on a real dataset and compare it with commonly used open API recommendation approaches: term frequency-inverse document frequency, latent dirichlet allocation, Word2Vec, and Sentence-BERT. The results show that the DLOAR approach has better performance than the other approaches in terms of precision, recall, F1-measure, mean average precision,and mean reciprocal rank.

关键词： Mashup development open API recommendation deep learning neural network service discovery

来源：评论

学校读者我要写书评

暂无评论

From black box AI to XAI in neuro-oncology: a survey on MRI-based tumor detection

引用

Discover Artificial Intelligence 2025年第1期5卷 1-21页

作者： Asmita Mittal, Praveen Department of Computer Engineering and Application GLA University Uttar Pradesh Mathura Bharthia India

Brain tumor detection remains a critical focus in neuro-oncology, requiring precise and efficient diagnostic methods to support timely clinical decisions. Magnetic Resonance Imaging (MRI) is the modality of choice for this task, owing to its exceptional soft tissue contrast and spatial resolution. Recent advancements in deep transfer learning have driven transformative progress in automating tumor detection, achieving notable improvements in accuracy, scalability, and computational efficiency. This paper systematically reviews state-of-the-art transfer learning techniques, including pre-trained neural networks, domain adaptation approaches, fine-tuning methodologies, and ensemble learning frameworks. It also provides an in-depth analysis of widely used MRI datasets, highlighting persistent challenges such as data imbalance, overfitting, and variability across imaging domains. Emerging trends in model optimization, architectural innovation, and generalization strategies are discussed, offering a forward-looking perspective on the future of automated tumor detection. Through a detailed comparative evaluation, this review identifies the strengths and limitations of existing methods, while outlining potential directions for research aimed at developing robust, interpretable, and clinically deployable systems for brain tumor diagnosis. © The Author(s) 2025.

关键词： Magnetic resonance imaging

来源：评论

学校读者我要写书评

暂无评论

Adaptive coupled-sliding-variable-based finite-time control of composite formation for multi-robot systems

引用

Science China(Information Sciences) 2024年第10期67卷 244-260页

作者： Xinru MA Hengyu LI Jun LIU Yueying WANG Shaorong XIE Jun LUO School of Mechatronic Engineering and Automation Shanghai University School of Mathematics and Computer Application Technology Jining University School of Computer Science Shanghai University College of Mechanical and Vehicle Engineering Chongqing University

This paper focuses on the finite-time control(FTC) of the composite formation consensus(CFC)problems for multi-robot systems(MRSs). The CFC problems are firstly proposed for MRSs under the complex network topology of cooperative or cooperative-competitive networks. Regarding the problems of FTC and CFC on multiple Lagrange systems(MLSs), coupled sliding variables are introduced to deal with the robustness and consistent convergence. Then, the adaptive finite-time protocols are given based on the displacement approaches. With the premised FTC, tender-tracking methods are further developed for the problems of tracking information disparity. Stability analyses of those MLSs mentioned above are clarified with Lyapunov candidates considering the coupled sliding vectors, which provide new verification for tender-tracking systems. Under the given coupled-sliding-variable-based finite-time protocols, MLSs distributively adjust the local formation error to achieve global CFC and perform uniform convergence in time-varying tracking. Finally, simulation experiments are conducted while providing practical solutions for the theoretical results.

关键词： composite formation time-vary tracking information disparity tender-tracking finite-time control nonlinear robots complex networks symmetry consensus

来源：评论

学校读者我要写书评

暂无评论

Automatic Image Acquisition and Matching in Control System of Near-Field Beam Shaping 8

Automatic Image Acquisition and Matching in Control System o...

引用

8th International Symposium on computer Science and Intelligent Control, ISCSIC 2024

作者： Li, Suicheng Shen, Wenwen He, Zizhou Zhou, Chen Li, Li Luo, Jun Institute of Computer Application Caep Department of Computer Engineering and Science Mianyang China

ISBN: (纸本)9798350380286

In the high power laser system of inertial Confinement fusion (ICF), in order to improve the safe operation flux of the system, near-field beam shaping is necessary to output the laser beam with uniform and smooth intensity distribution. Therefore, how to use optical control systems to automatically and efficiently accomplish near-field beamforming is an important issue. The optical control system based on liquid crystal spatial light modulator can control the beam shape and intensity distribution in real time. The control system of near-field beam shaping algorithm mainly includes the modules of light path control, image loading on liquid crystal spatial light modulator, Charge Coupled Device (CCD) nearfield light spot perception and image processing. There is a lot of data exchange between those modules. In the laser system, the manual operation mainly involves two parts: one is the interaction and data exchange between different modules. In the past work, the image data is collected by the operator on different devices, and then the compensation image is calculated on the computing device, and finally loaded into the display device. The whole process is complex, tedious and repetitive. The other is the image matching of input image on spatial light modulator and near-field light spot. This work requires staff to manually match feature points on computing equipment, and the accuracy of matching is difficult to guarantee, which is very time-consuming and labor-intensive. This paper will address the above problems and propose a distributed network architecture to connect different devices in near-field beam shaping and carry out image acquisition from them. In addition, a new image matching algorithm is proposed to extract and match the feature points of input image on spatial light modulator and near-field light spot. It enables the compensation algorithm to be completed automatically to avoid manual operation. © 2024 IEEE.

关键词： Light modulators

来源：评论

学校读者我要写书评

暂无评论

Deep Learning Used for Recognition of Landmark 4

Deep Learning Used for Recognition of Landmark

引用

4th IEEE International Conference on Data engineering and Communication Systems, ICDECS 2024

作者： Prasad, Arpana Asha, V. Sundar, Chirala Shanmukha Guddenna, Divyabharathi T Gaganashree, C. New Horizon College of Engineering Computer Application Bengaluru India

ISBN: (纸本)9798350393354

As Google has made its large-scale dataset available to the public, innovation in the field of Google Landmark Classifications has taken a forefront for research in field of Deep Learning. The Google dataset contains photos of different landmarks aggregated over a period of time. The contributions are made by millions of people for thousands of monuments worldwide. The dataset is large and is unbalanced posing problems to researchers for building models. This study demonstrates the application of transfer learning in conjunction with data augmentation to produce a model that yields 82.03% Top-5 accuracy on a modified version of the original Google-Landmarks dataset, which comprises pictures from 6,151 distinct landmarks. We also discuss the application of a Generative Adversarial Network in mitigating the issues caused by an exceedingly unbalanced © 2024 IEEE.

关键词： Classification (of information)

来源：评论

学校读者我要写书评

暂无评论

Genetic Algorithm–Aided Deep Feature Selection for Improved Rice Disease Classification

引用

Operations Research Forum 2025年第1期6卷 1-24页

作者： Sharma, Rahul Singh, Amar Kumar, Prashant Singh, Mahipal Department of Computer Application GDC Marh Jammu India School of Computer Application Lovely Professional University Phagwara India School of Mechanical Engineering Lovely Professional University Phagwara India

Pests and diseases pose significant threats to crop safety and accessibility. Deep learning integration with traditional pest management fosters sustainable agriculture, minimizes chemical pesticide use, and facilitates early detection of pests and crop diseases through image and sensor data analysis. To ensure food security, reduce costs, and enhance overall production, computer vision techniques are essential for processing complex, high-dimensional real-world data. However, data dimensionality reduction is crucial for achieving accurate disease identification. Pre-trained models can extract valuable features from data. In this study, InceptionV3 and MobileNet V2 were utilized to extract comprehensive features from a small UCI rice leaf disease dataset. However, classifier performance using these features is uncertain due to factors such as domain mismatch, high-level data representation, model biases, dataset variability, and task complexity. To address these challenges, a rigorous evaluation of pre-trained feature suitability is necessary. A genetic algorithm–based feature selection (FS) approach was employed. FS streamlines data, reducing the information required for disease identification. An optimized ANN classifier, MobileNet-GA-ANN, achieved a 96.58% accuracy, outperforming other methods. © The Author(s), under exclusive licence to Springer Nature Switzerland AG 2024.

关键词： Feature selection Genetic algorithm Plant disease detection Soft computing Transfer learning

来源：评论

学校读者我要写书评

暂无评论

Integrated fuzzy-based gamma adaptive for image enhancement with color attenuation prior

引用

Multimedia Tools and applications 2025年第12期84卷 11249-11267页

作者： Harne, Sachin Choubey, Siddhartha Choubey, Abha Department of Computer Application Shri Shankaracharya Group of Institution Chhattisgarh Bhilai India Department of Computer Science and Engineering Shri Shankaracharya Group of Institution Chhattisgarh Bhilai India

Images Quality in the Transportation industry is considered the most significant term and gains more attention among the higher authorities and researchers. Therefore, the accuracy level of the image has been evaluated with IQA (Image Quality Assessment). Most images are substantially affected by ambient lights that affect the quality of the image. In the existing research, the HVS (Human Visual System) has been utilized in order to identify the quality of images that occurred in both night and day time that affects the loss of luminance and information. In the current research, we proposed an integrated fuzzy-based Gamma Adaptive enhanced with Color Attenuation prior (IFGECAP) techniques that help capture the image's local characteristics. Eventually, the performance evaluation of the image whereas both input and output, has been evaluated with Natural image quality evaluator (NIQE), Blind/Reference less Image Spatial Quality Evaluator (BRISQUE), structural similarity index measure (SSIM), Signal-to-noise ratio (SNR), and Peak signal-to-noise ratio (PSNR). Hence, the fog removal will be classified as seasonal, such as rainy and foggy, utilizing CNN—Convolutional Neural Network. The proposed hybrid technique—IFGECAP, procured an efficient result when compared with the existing techniques. © The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature 2024.

关键词： Signal to noise ratio

来源：评论

学校读者我要写书评

暂无评论

Proposing a digital twin-based sustainable water governance system for rural Indian villages

引用

International Journal of Information Technology (Singapore) 2025年第3期17卷 1777-1783页

作者： Singh, Dineshkumar Sharma, Vishnu Department of Computer Science and Engineering Faculty of Computer Science & Engineering (FCE) Poornima University Jaipur India Department of Computer Science and Application Faculty of Computer Science & Engineering (FCE) Poornima University Jaipur India

A sustainably governed water-ecosystem at village-level is crucial for the community's well-being. It requires understanding natures’ limits to store and yield water and balance it with the stakeholders’ needs, in the context of population growth, climate change, and unsustainable agricultural practices. Maharashtra villages like Randullabad and Karanjkhop are impacted by limited groundwater availability due to the geological structure of the Deccan Traps basaltic rock formation and are recharged by the most rainfall during the monsoon season. Traditional water management practices are inadequate, with old wells causing conflicts among villagers due to fluctuating water levels. In order to achieve a sustainable water system for rural Indian villages, this paper proposes a Digital Twin-Based Water Governance System (DToWGS) with the aim of balancing nature's boundaries with stakeholders' interests to provide real-time water insights, encourage community engagement, transparency in water allocation and decision-making, and promote sustainable practices. These measured indicators can potentially link local governance outcomes to global goals like Viksit Bharat 2047 and UN Sustainable Development Goal 6 (SDG 6). © Bharati Vidyapeeth's Institute of computer applications and Management 2025.

关键词： Digital twin Empathy map Remote sensing SDG6 Stakeholders Village water environment

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：