检索结果-内蒙古大学图书馆

Valuing Training Data via Causal Inference for In-Context Learning

IEEE Transactions on Knowledge and Data engineering 2025年第6期37卷 3824-3840页

作者： Zhou, Xiaoling Ye, Wei Lee, Zhemg Zou, Lei Zhang, Shikun Peking University National Engineering Research Center for Software Engineering Beijing100871 China Tianjin University Tianjin300072 China Peking University Wangxuan Institute of Computer Technology Beijing100871 China

In-context learning (ICL) empowers large pre-trained language models (PLMs) to predict outcomes for unseen inputs without parameter updates. However, the efficacy of ICL heavily relies on the choice of demonstration examples. Randomly selecting from the training set frequently leads to inconsistent performance. Addressing this challenge, this study takes a novel approach by focusing on training data valuation through causal inference. Specifically, we introduce the concept of average marginal effect (AME) to quantify the contribution of individual training samples to ICL performance, encompassing both its generalization and robustness. Drawing inspiration from multiple treatment effects and randomized experiments, we initially sample diverse training subsets to construct prompts and evaluate the ICL performance based on these prompts. Subsequently, we employ Elastic Net regression to collectively estimate the AME values for all training data, considering subset compositions and inference performance. Ultimately, we prioritize samples with the highest values to prompt the inference of the test data. Across various tasks and with seven PLMs ranging in size from 0.8B to 33B, our approach consistently achieves state-of-the-art performance. Particularly, it outperforms Vanilla ICL and the best-performing baseline by an average of 14.1% and 5.2%, respectively. Moreover, prioritizing the most valuable samples for prompting leads to a significant enhancement in performance stability and robustness across various learning scenarios. Impressively, the valuable samples exhibit transferability across diverse PLMs and generalize well to out-of-distribution tasks. © 1989-2012 IEEE.

关键词： Contrastive Learning

来源：评论

学校读者我要写书评

暂无评论

Observability in Microservices: An In-Depth Exploration of Frameworks, Challenges, and Deployment Paradigms

引用

IEEE Access 2025年 13卷 72011-72039页

作者： Faseeha, Ummay Jamil Syed, Hassan Samad, Fahad Zehra, Sehar Ahmed, Hamza Department of Computer Science Karachi75030 Pakistan Jinnah University for Women Department of Computer Science and Software Engineering Karachi74600 Pakistan Kuala Lumpur57000 Malaysia Department of Cyber Security Karachi75030 Pakistan

The widespread adoption of microservices-based architectures in native cloud systems has amplified the need for robust observability strategies to ensure system reliability and performance. Microservices, while enabling scalability and flexibility, introduce complexities such as distributed failure points, performance bottlenecks, and resource mismanagement. This paper provides a comprehensive review of state-of-the-art observability frameworks and tools designed for microservices architecture. It focuses on key aspects of observability at the system, service, and network levels, utilizing logs, traces, and metrics as core data sources. A thematic taxonomy is proposed, classifying the diverse approaches to monitoring microservices based on implementation paradigms, deployment platforms, and architecture patterns. The study compares existing solutions in terms of their capabilities, limitations, and effectiveness in addressing observability challenges in cloud, edge, and fog environments. Furthermore, open research issues are identified, and practical recommendations are provided to optimize observability in complex, distributed microservices ecosystems. This work aims to serve as a valuable resource for system architects, DevOps engineers, and researchers striving to enhance the reliability and performance of modern cloud-native systems. © 2013 IEEE.

关键词： Cloud platforms

来源：评论

学校读者我要写书评

暂无评论

UEFL: Universal and Efficient Privacy-Preserving Federated Learning

引用

IEEE Internet of Things Journal 2025年第10期12卷 14333-14347页

作者： Li, Zhiqiang Bao, Haiyong Pan, Hao Guan, Menghong Huang, Cheng Dai, Hong-Ning East China Normal University Software Engineering Institute Shanghai200062 China Fudan University School of Computer Science Shanghai200438 China Hong Kong Baptist University Department of Computer Science Hong Kong

Federated Learning (FL) is a distributed machine learning framework that allows for model training across multiple clients without requiring access to their local data. However, FL poses some risks, for example, curious clients might conduct inference attacks (e.g., membership inference attacks, model-inversion attacks) to extract sensitive information from other participants. Existing solutions typically fail to strike a good balance between performance and privacy, or are only applicable to specific FL scenarios. To address these challenges, we propose a universal and efficient privacy-preserving FL framework based on matrix theory. Specifically, we design the Improved Extended Hill Cryptosystem (IEHC), which efficiently encrypts model parameters while supporting the secure ReLU function. To accommodate different training tasks, we design the Secure Loss Function Computation (SLFC) protocol, which computes derivatives of various loss functions while maintaining data privacy of both client and server. And we implement SLFC specifically for three classic loss functions, including MSE, Cross Entropy, and L1. Extensive experimental results demonstrate that our approach robustly defends against various inference attacks. Furthermore, model training experiments conducted in various FL scenarios indicate that our method shows significant advantages across most metrics. © 2014 IEEE.

关键词： Federated learning

来源：评论

学校读者我要写书评

暂无评论

Two-Timescale Hierarchical Contract for Joint Computation Offloading and Energy Management in Edge Computing System

引用

IEEE Transactions on Network Science and engineering 2025年第3期12卷 1745-1760页

作者： Yan, Min Wang, Li Xu, Lianming Hou, Luyang Han, Zhu Beijing University of Posts and Telecommunications School of Computer Science National Pilot Software Engineering School Beijing100876 China Beijing University of Posts and Telecommunications School of Electronic Engineering Beijing100876 China University of Houston Department of Electrical and Computer Engineering HoustonTX77004 United States Kyung Hee University Department of Computer Science and Engineering Seoul446-701 Korea Republic of

To mitigate the rising energy costs in edge computing, edge servers (ESs) can receive revenues from reducing their energy usage by contracting with virtual power plant (VPP). ESs also respond to user equipment (UE) by providing computation offloading services. However, such two-layer coordinated trading of computation offloading and energy management in a VPP-ES-UE architecture should address the information asymmetry issue and uncertain task arrivals, posing challenges to maximizing the stakeholders' utility. In this paper, we formulate the two-layer coordinated trading as a hierarchical contracting problem, which addresses the information asymmetry using two contract models. We design a VPP-ES energy contract on a large timescale and an ES-UE computation contract on a small timescale, where ESs need to coordinate the future task arrivals under the energy reduction target assigned by VPP. We construct a two-timescale virtual queue to achieve the energy reduction goal by the long-term queue stability constraint and employ Lyapunov optimization to transform the original problem into an online optimization problem without requiring future information. An online two-timescale hierarchical contract optimization algorithm is proposed to solve the transformed problem. The simulation results demonstrate that our method achieves higher social welfare compared to other benchmarks. © 2013 IEEE.

关键词： Computation offloading

来源：评论

学校读者我要写书评

暂无评论

When Crypto Fails: Demystifying Cryptographic Defects in Ethereum Smart Contracts

引用

IEEE Transactions on software engineering 2025年第05期51卷 1381-1398页

作者： Zhang, Jiashuo Chen, Jiachi Shen, Yiming Zhang, Tao Wang, Yanlin Chen, Ting Gao, Jianbo Chen, Zhong Peking University School of Computer Science China Sun Yat-sen University School of Software Engineering China Macau University of Science and Technology School of Computer Science and Engineering China University of Electronic Science and Technology of China School of Computer Science and Engineering China Beijing Jiaotong University Beijing Key Laboratory of Security and Privacy in Intelligent Transportation China

Ethereum has officially provided a set of system-level cryptographic APIs to enhance smart contracts with cryptographic capabilities. These APIs have been utilized in over 13.8% of Ethereum transactions, motivating developers to implement various on-chain cryptographic tasks, such as digital signatures. However, since developers may not always be cryptographic experts, their ad-hoc and potentially defective implementations could compromise the theoretical guarantees of cryptography, leading to real-world security issues. To mitigate this threat, we conducted a comprehensive study aimed at demystifying and detecting cryptographic defects in smart contracts. Through the analysis of 3,762 real-world security reports, we defined 12 types of cryptographic defects in smart contracts with detailed descriptions and practical detection patterns. Based on this categorization, we proposed CryptoScan, the first static analyzer to automate the pre-deployment detection of cryptographic defects in smart contracts. CryptoScan utilizes cross-contract and inter-procedure static analysis to identify crypto-related execution paths and employs taint analysis to extract fine-grained crypto-specific semantics for defect detection. Furthermore, we collected a large-scale dataset containing 79,598 real-world crypto-related smart contracts and evaluated CRYPTOSCAN’s effectiveness on it. The results demonstrated that CryptoScan achieves an overall precision of 96.1% and a recall of 93.3%. Notably, CryptoScan revealed that 19,707 (24.8%) out of 79,598 smart contracts contain at least one cryptographic defect. Although not all defects directly cause financial losses, they indicate prevalent non-standard cryptographic implementations that should be addressed in real-world practices. © 2025 IEEE. All rights reserved.

关键词： Smart contract

来源：评论

学校读者我要写书评

暂无评论

Priority-Aware Resource Allocation for VNF Deployment in Service Function Chains Based on Graph Reinforcement Learning

引用

computers, Materials & Continua 2025年第5期83卷 1649-1665页

作者： Seyha Ros Seungwoo Kang Taikuong Iv Inseok Song Prohim Tam Seokhoon Kim Department of Software Convergence Soonchunhyang UniversityAsan31538Republic of Korea School of Digital Technologies American University of Phnom PenhPhnom Penh12106Cambodia Department of Computer Software Engineering Soonchunhyang UniversityAsan31538Republic of Korea

Recently,Network Functions Virtualization(NFV)has become a critical resource for optimizing capability utilization in the 5G/B5G *** decomposes the network resource paradigm,demonstrating the efficient utilization of Network Functions(NFs)to enable configurable service priorities and resource *** Service Providers(TSPs)face challenges in network utilization,as the vast amounts of data generated by the Internet of Things(IoT)overwhelm existing *** applications,which generate massive volumes of diverse data and require real-time communication,contribute to bottlenecks and *** this context,Multiaccess Edge Computing(MEC)is employed to support resource and priority-aware IoT applications by implementing Virtual Network Function(VNF)sequences within Service Function Chaining(SFC).This paper proposes the use of Deep Reinforcement Learning(DRL)combined with Graph Neural Networks(GNN)to enhance network processing,performance,and resource pooling *** facilitates feature extraction through Message-Passing Neural Network(MPNN)*** with DRL,Deep Q-Networks(DQN)are utilized to dynamically allocate resources based on IoT network priorities and *** focus is on minimizing delay times for VNF instance execution,ensuring effective resource placement,and allocation in SFC deployments,offering flexibility to adapt to real-time changes in priority and *** results demonstrate that our proposed scheme outperforms reference models in terms of reward,delay,delivery,service drop ratios,and average completion ratios,proving its potential for IoT applications.

关键词： Deep reinforcement learning graph neural network multi-access edge computing network functions virtualization software-defined networking

来源：评论

学校读者我要写书评

暂无评论

IEEE Transactions on Artificial Intelligence

引用

IEEE Transactions on Artificial Intelligence 2025年第5期6卷 1360-1372页

作者： Wang, Shuyue Liu, Zhunga Zhang, Zuowei Bennamoun, Mohammed Northwestern Polytechnical University School of Automation Shaanxi Xi’an710072 China The University of Western Australia Department of Computer Science and Software Engineering PerthWA6009 Australia

Multimodal domain adaptation (MMDA) aims to transfer knowledge across different domains that contain multimodal data. Current methods typically assume that both the source and target domains have paired multimodal data with the same modalities, allowing for direct knowledge transfer between corresponding types of data. However, in certain applications, the source domain benefits from advanced sensors and equipment, capturing more modalities than those available in the target domain. As a result, the information from the source modalities may not strongly align with that of the target modalities. This weak correlation hinders the effective utilization of all source data for the target domain. To address this challenge, we propose a weakly correlated multimodal domain adaptation (WCMMDA) method for pattern classification. WCMMDA is designed to acquire the modality-independent and category-related knowledge from the source domain, enabling the full utilization of available source modalities for effective knowledge transfer. Specifically, modality-invariant features are first extracted from the multimodal data to bridge the heterogeneity gap within each domain. Subsequently, domain-invariant features are further learned from these modality-invariant features to align the feature distributions across the source and target domains. A source-specific classifier is employed here, which predicts pseudo-labels for the target data and enables the feature extractor to explore category-related information in source features. Finally, a target-specific classifier is trained using the pseudolabeled target data, where highly reliable pseudolabels are selected based on confidence to improve classification performance. Extensive experiments are performed on the real-world multimodal datasets to demonstrate the superiority of WCMMDA. © 2020 IEEE.

关键词： Labeled data

来源：评论

学校读者我要写书评

暂无评论

An Empirical Study on Small-Sized Datasets Based on Eubank’s Optimal Spacing Theorem

引用

SN computer Science 2025年第1期6卷 1-19页

作者： Abedu, Samuel Mensah, Solomon Boafo, Frederick Department of Computer Science and Software Engineering Concordia University Montreal Canada Department of Computer Science University of Ghana Accra Ghana Department of Computer Science Lancaster University Ghana Accra Ghana

Conventional machine learning methods for software effort estimation (SEE) have seen an increase in research interest. Conversely, there are few research that try to evaluate how well deep learning techniques work in SEE. This can be attributed to the relatively small sizes of SEE datasets. The goal of the study is to establish a threshold for small-sized datasets in SEE. Additionally, it looks into how well certain deep learning and traditional machine learning models perform on small-sized datasets. From the body of existing literature, plausible SEE datasets are extracted and ranked along with their attributes and number of project cases. The ranking of the project instances is discretized into three classes (small, medium, and large) using Eubank’s optimal spacing theory. Using the leave-one-out cross-validation, each small-sized dataset is used to train two deep learning models and five conventional machine learning models. Each model’s ability to make predictions is evaluated using its mean absolute error. Results show that, on small-scale datasets, deep learning models outperform traditional machine learning models in terms of prediction accuracy, which contradicts what is previously known. Regularisation techniques can be used in conjunction with deep learning to address SEE. © The Author(s), under exclusive licence to Springer Nature Singapore Pte Ltd. 2024.

关键词： Deep learning Eubank’s optimal spacing theory Small-sized software effort estimation Traditional Machine learning

来源：评论

学校读者我要写书评

暂无评论

Road Defect Detection Model Based on YOLOv8

IAENG International Journal of Computer Science

引用

IAENG International Journal of computer Science 2025年第4期52卷 986-994页

作者： Cong, Shengqiang Zhang, Chunna Yu, Yang Yue, Xiaoping Zhao, Jinchi Shen, Yuming School of Computer Science and Software Engineering University of Science and Technology Liaoning Liaoning Anshan114051 China University of Science and Technology Liaoning Liaoning Anshan114051 China

The detection of road defects is crucial for ensuring vehicular safety and facilitating the prompt repair of roadway imperfections. Existing YOLOv8-based models face the following issues: extraction capabilities and insufficient feature representation in complex scenarios, slow bounding box regression speed. To address these challenges, we propose the YOLOv8-DSW model. Firstly, we incorporate the Dilation-wise Residual (DWR) module into the C2f module in the neck, improving the detection performance of multi-scale feature maps. Secondly, We incorporate the SENetv2 attention mechanism into the neck of the YOLOv8 model to augment feature expression capabilities. Finally, we introduce Wise-IoU (WIoU) to enhance the speed of bounding box regression. Experimental results indicate that YOLOv8-DSW enhances the mean Average Precision at 50% (mAP50) by 2.5% in comparison to the YOLOv8n, significantly improving detection accuracy and provide valuable method for the development of road defect detection. © (2025.0), (International Association of Engineers). All Rights Reserved.

关键词： Deep learning

来源：评论

学校读者我要写书评

暂无评论

SQUMUTH squirrel search based algorithm for high order mutant generation in mutation testing

引用

Discover Computing 2025年第1期28卷 1-17页

作者： Mohanty, Subhasish Mishra, Jyotirmaya Mohapatra, Sudhir Kumar Bejo, Seifu Detso Deferisha, Aliazar Deneke Computer Science and Engineering GIET University Odisha Gunupur India Faculty of Engineering and Technology Sri Sri University Cuttack India Department of Software Engineering CCI Wolkite University Wolkite Ethiopia Faculty of Computing and Software Engineering Arba Minch University Arba Minch Ethiopia

In today's software testing community, quality assessment remains critical, with mutation testing standing as a cornerstone technique for evaluating the effectiveness of test cases. This method involves introducing faulty code, or mutants, into the program to assess the quality of the test suite and other testing methods. However, mutation testing faces challenges such as the generation of numerous mutants, the presence of equivalent mutants that are difficult to detect through testing, and the lack of realistic mutant creation. Literature reviews indicate significant efforts to address these issues through formal solutions and heuristic methods. Recent optimization-based methods are now recognized as cost-effective result in an optimized solution. Hence, to overcome these limitations, this study introduces SQUMUTH, a novel approach for high-order mutant generation based on the Squirrel Search Algorithm (SSA). Inspired by the foraging behavior of squirrels, SSA offers a promising solution for enhancing the efficiency and effectiveness of mutation testing. Experimental evaluations on eight well-known Java benchmark programs demonstrate that SQUMUTH outperforms existing methods. Comparative analyses of mutation scores and the rates of realistic mutants consistently show its better performance across all subject programs compared to other state-of-the-art methods such as Social Group Optimization, Binary Genetic Algorithm, and random testing. The experimental results underscore its effectiveness in generating more realistic mutants. The experimental results indicated that the proposed approach has the potential to advance software testing by improving the cost-effectiveness of mutation analysis and the quality of software systems. © The Author(s) 2025.

关键词： First-order mutants Higher-order mutants Mutation score Mutation testing software testing Squirrel search algorithm

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：