检索结果-内蒙古大学图书馆

2025 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2025

作者： Tao, Zhen Cao, Yuehang Fang, Yang Liu, Yunhui Zhao, Xiang He, Tieke State Key Laboratory for Novel Software Technology Nanjing University Nanjing China Laboratory for Big Data and Decision National University of Defense Technology Changsha China National University of Defense Technology Changsha China

ISBN: (纸本)9798350368741

Dynamic recommendation, focusing on modeling user preference from historical interactions and providing recommendations on current time, plays a key role in many personalized services. Recent works show that pre-trained dynamic graph neural networks (GNNs) can achieve excellent performance. However, existing methods by fine-tuning node representations at large scales demand significant computational resources. Additionally, the long-tail distribution of degrees leads to insufficient representations for nodes with sparse interactions, posing challenges for efficient fine-tuning. To address these issues, we introduce GraphSASA, a novel method for efficient fine-tuning in dynamic recommendation systems. GraphSASA employs test-time augmentation by leveraging the similarity of node representation distributions during hierarchical graph aggregation, which enhances node representations. Then it applies singular value decomposition, freezing the original vector matrix while focusing fine-tuning on the derived singular value matrices, which reduces the parameter burden of fine-tuning and improves the fine-tuning adaptability. Experimental results demonstrate that our method achieves state-of-the-art performance on three large-scale datasets. © 2025 IEEE.

关键词： graph neural networks low-rank adaptation pre-training recommendation

来源：评论

学校读者我要写书评

暂无评论

A Comprehensive Review on Deep Learning System Testing 24th

A Comprehensive Review on Deep Learning System Testing

引用

24th International Conference on Algorithms and Architectures for Parallel Processing, ICA3PP 2024

作者： Li, Ying Shan, Chun Liu, Zhen Liao, Shuyan Beijing Key Laboratory of Software Security Engineering Technology Beijing Institute of Technology Beijing China School of Computer Science and Technology Beijing Institute of Technology Beijing China Computer School Beijing Information Science and Technology University Beijing China School of Cyberspace Science and Technology Beijing Institute of Technology Beijing China

ISBN: (纸本)9789819615476

Deep learning(DL) systems exhibit multiple behavioral characteristics such as correctness, robustness, and fairness. Ensuring that these behavioral characteristics function properly is crucial for maintaining the accuracy of DL systems’ outputs. As a specialized form of software, DL systems’ security testing techniques have increasingly become a focus of research in quality assurance. We analyze and organize the testing techniques for DL systems based on an investigation of the current state of the art both domestically and internationally. This paper categorizes existing approaches as component-oriented and attribute-oriented methods, providing a detailed review based on this classification. Additionally, we forecast the future development of testing techniques for DL systems. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025.

关键词： Adversarial machine learning

来源：评论

学校读者我要写书评

暂无评论

PFedCS: A Personalized Federated Learning Method for Enhancing Collaboration among Similar Classifiers 39

PFedCS: A Personalized Federated Learning Method for Enhanci...

引用

39th Annual AAAI Conference on Artificial Intelligence, AAAI 2025

作者： Wu, Siyuan Jia, Yongzhe Liu, Bowen Xiang, Haolong Xu, Xiaolong Dou, Wanchun State Key Laboratory for Novel Software Technology School of Computer Science Nanjing University China School of Software Nanjing University of Information Science and Technology China

ISBN: (纸本)157735897X

Personalized federated learning (PFL) has recently gained significant attention for its capability to address the poor convergence performance on highly heterogeneous data and the lack of personalized solutions of traditional federated learning (FL). Existing mainstream approaches either perform personalized aggregation based on a specific model architecture to leverage global knowledge or achieve personalization by exploiting client similarities. However, the former overlooks the discrepancies in client data distributions by indiscriminately aggregating all clients, while the latter lacks fine-grained collaboration of classifiers relevant to local tasks. In view of this challenge, we propose a Personalized Federated learning method for Enhancing Collaboration among Similar Classifiers (PFedCS), which aims at improving the client’s accuracy on local tasks. Concretely, it is achieved by leveraging awareness of the client classifier similarities to address the above problems. By iteratively measuring the distance of the classifier parameters between clients and clustering with each client as a cluster center, the central server adaptively identifies the collaborating clients with similar data distributions. In addition, a distance-constrained aggregation method is designed to generate customized collaborative classifiers to guide local training. As a result, extensive experimental evaluations conducted on various datasets demonstrate that our method achieves state-of-the-art performance. © 2025, Association for the Advancement of Artificial Intelligence (***). 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

TH-SLP: Web Service Link Prediction Based on Topic-aware Heterogeneous Graph Neural Network

TH-SLP: Web Service Link Prediction Based on Topic-aware Het...

引用

2023 IEEE International Conference on Web Services, ICWS 2023

作者： Peng, Qian Cao, Buqing Xie, Xiang Liu, Shanpeng Kang, Guosheng Liu, Jianxun Hunan University of Science and Technology Xiangtan School of Computer Science and Engineering Hunan Provincial Key Lab. for Services Computing and Novel Software Technology Hunan China

ISBN: (纸本)9798350304855

With the emergence of more and more Web services, finding suitable services becomes a difficult problem. Service link prediction is employed to disclose relationships among services, which facilitates the further development of service composition, selection, and recommendation. But the existing link prediction approaches simply utilize the structural features of the service network. In reality, the rich text content in service node description documents also carries latent but fine-grained semantics generated by multifaceted topic-aware factors, yet few efforts are committed to mining them. In this paper, we propose a Web service link prediction method based on a topic-aware heterogeneous graph neural network. Specifically, the method consists of two main layers, including the meta-path intra-decomposition and the meta-path inter-mergence. Meta-path intra-decomposition aims to mine the topic distribution of the meta-paths-based context while capturing fine-grained topic-aware semantics. Meta-path inter-mergence uniquely aggregates topic-aware factors according to the mined distribution and adopts a multifaceted attention mechanism to aggregate different meta-paths, enabling service nodes to generate multifaceted topic-aware embeddings that preserve not only the structure and but also the topic-aware semantics. In addition, a topic prior guidance regularization item is set up for quality assurance of multifaceted topic-aware embedding that depends on global knowledge of the unstructured text content in description documents. Experimental results on real datasets show that our proposed model outperforms other existing baselines methods in the link prediction task, successfully validating the effectiveness of our proposed method. © 2023 IEEE.

关键词： Semantics

来源：评论

学校读者我要写书评

暂无评论

Boosting Adversarial Training with Learnable Distribution

引用

computers, Materials & Continua 2024年第3期78卷 3247-3265页

作者： Kai Chen Jinwei Wang James Msughter Adeke Guangjie Liu Yuewei Dai School of Electronics and Information Engineering Nanjing University of Information Science and TechnologyNanjing210044China Key Laboratory of Intelligent Support Technology for Complex Environments Ministry of EducationNanjing210044China School of Computer and Software Nanjing University of Information Science and TechnologyNanjing210044China Nanjing Center for Applied Mathematics Nanjing211135China

In recent years,various adversarial defense methods have been proposed to improve the robustness of deep neural *** training is one of the most potent methods to defend against adversarial ***,the difference in the feature space between natural and adversarial examples hinders the accuracy and robustness of the model in adversarial *** paper proposes a learnable distribution adversarial training method,aiming to construct the same distribution for training data utilizing the Gaussian mixture *** distribution centroid is built to classify samples and constrain the distribution of the sample *** natural and adversarial examples are pushed to the same distribution centroid to improve the accuracy and robustness of the *** proposed method generates adversarial examples to close the distribution gap between the natural and adversarial examples through an attack algorithm explicitly designed for adversarial *** algorithm gradually increases the accuracy and robustness of the model by scaling ***,the proposed method outputs the predicted labels and the distance between the sample and the distribution *** distribution characteristics of the samples can be utilized to detect adversarial cases that can potentially evade the model *** effectiveness of the proposed method is demonstrated through comprehensive experiments.

关键词： Adversarial training feature space learnable distribution distribution centroid

来源：评论

学校读者我要写书评

暂无评论

Raven: Benchmarking Monetary Expense and Query Efficiency of OLAP Engines on the Cloud 28th

Raven: Benchmarking Monetary Expense and Query Efficiency o...

引用

28th International Conference on Database Systems for Advanced Applications, DASFAA 2023

作者： Wu, Tongyu Gu, Rong Li, Yang Ma, Hongbin Chen, Yi Zhu, Ying Yu, Xiaoxiang Xu, Tengting Huang, Yihua State Key Laboratory for Novel Software Technology Nanjing University Nanjing China Kyligence Inc. Shanghai China

ISBN: (纸本)9783031306778

Nowadays, it is prevalent to build OLAP services on cloud platforms. Cloud OLAP adopters are eager to understand and characterize the performance of OLAP engines on the cloud. However, traditional OLAP benchmarks are usually designed for on-premise environments. When evaluating cloud OLAP engines, they have limitations on cloud environment adaption and cloud scenario benchmark execution. To address these issues, this paper proposes Raven, a cloud-oriented OLAP benchmark with flexible system architecture and diversified workloads. Raven supports cloud service deployment and various cloud OLAP engine integration. In addition, to simulate complex cloud query scenarios, we design a group of timeline-based and service-oriented workloads. We implement Raven on the Amazon AWS cloud platform and use it to evaluate typical types of widely-used OLAP engines, including Presto, SparkSQL, Kylin, and Athena. Experimental results show that Raven can effectively benchmark diversified OLAP engines. Besides, Raven can benchmark various configuration settings of an identical OLAP engine. We also explore an OLAP case study on the cloud using Raven. © 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.

关键词： Engines

来源：评论

学校读者我要写书评

暂无评论

LIBALCHEMY: A Two-Layer Persistent Summary Design for Taming Third-Party Libraries in Static Bug-Finding Systems 24

LIBALCHEMY: A Two-Layer Persistent Summary Design for Taming...

引用

44th ACM/IEEE International Conference on software Engineering, ICSE 2024

作者： Wu, Rongxin He, Yuxuan Huang, Jiafeng Wang, Chengpeng Tang, Wensheng Shi, Qingkai Xiao, Xiao Zhang, Charles Xiamen Key Laboratory of Intelligent Storage and Computing School of Informatics Xiamen University Xiamen China The Hong Kong University of Science and Technology Hong Kong State Key Laboratory for Novel Software Technology Nanjing University Nanjing China Person Hang Zhou China

ISBN: (纸本)9798400702174

Despite the benefits of using third-party libraries (TPLs), the misuse of TPL functions raises quality and security concerns. Using traditional static analysis to detect bugs caused by TPL function is nontrivial. One promising solution would be to automatically generate and persist the summaries of TPL functions offline and then reuse these summaries in compositional static analysis online. However, when dealing with millions of lines of TPL code, the summaries designed by existing studies suffer from an unresolved paradox. That is, a highly precise form of summary leads to an unaffordable space and time overhead, while an imprecise one seriously hurts its precision or recall. To address the paradox, we propose a novel two-layer summary design. The first layer utilizes a line-sized program representation known as the program dependence graph to compactly encode path conditions, while the second layer encodes bug-type-specific properties. We implemented our idea as a tool called Libalchemy and evaluated it on fifteen mature and extensively checked open-source projects. Experimental results show that Libalchemy can check over ten million lines of code within ten hours. Libalchemy has detected 55 true bugs with a high precision of 90.16%, eleven of which have been assigned CVE IDs. Compared to whole-program analysis and the conventional design of path-sensitively precise summaries, Libalchemy achieves an 18.56× and 12.77× speedup and saves 91.49% and 90.51% of memory usage, respectively. © 2024 ACM.

关键词： Static analysis

来源：评论

学校读者我要写书评

暂无评论

Long-Tail Learning with Foundation Model: Heavy Fine-Tuning Hurts 41

Long-Tail Learning with Foundation Model: Heavy Fine-Tuning ...

引用

41st International Conference on Machine Learning, ICML 2024

作者： Shi, Jiang-Xin Wei, Tong Zhou, Zhi Shao, Jie-Jing Han, Xin-Yan Li, Yu-Feng National Key Laboratory for Novel Software Technology Nanjing University China School of Artificial Intelligence Nanjing University China School of Computer Science and Engineering Southeast University China Key Laboratory of Computer Network and Information Integration Southeast University Ministry of Education China

The fine-tuning paradigm in addressing long-tail learning tasks has sparked significant interest since the emergence of foundation models. Nonetheless, how fine-tuning impacts performance in long-tail learning was not explicitly quantified. In this paper, we disclose that heavy fine-tuning may even lead to non-negligible performance deterioration on tail classes, and lightweight fine-tuning is more effective. The reason is attributed to inconsistent class conditions caused by heavy fine-tuning. With the observation above, we develop a low-complexity and accurate long-tail learning algorithms LIFT with the goal of facilitating fast prediction and compact models by adaptive lightweight fine-tuning. Experiments clearly verify that both the training time and the learned parameters are significantly reduced with more accurate predictive performance compared with state-of-the-art approaches. The implementation code is available at https://***/shijxcs/LIFT. Copyright 2024 by the author(s)

关键词： Contrastive Learning

来源：评论

学校读者我要写书评

暂无评论

SLAM: Towards Efficient Multilingual Reasoning via Selective Language Alignment 31

SLAM: Towards Efficient Multilingual Reasoning via Selective...

引用

31st International Conference on Computational Linguistics, COLING 2025

作者： Fan, Yuchun Mu, Yongyu Wang, Yilin Huang, Lei Ruan, Junhao Li, Bei Xiao, Tong Huang, Shujian Feng, Xiaocheng Zhu, Jingbo NLP Lab School of Computer Science and Engineering Northeastern University Shenyang China NiuTrans Research Shenyang China Harbin Institute of Technology Harbin China Meituan Inc. National Key Laboratory for Novel Software Technology Nanjing University China

ISBN: (纸本)9798891761964

Despite the significant improvements achieved by large language models (LLMs) in English reasoning tasks, these models continue to struggle with multilingual reasoning. Recent studies leverage a full-parameter and two-stage training paradigm to teach models to first understand non-English questions and then reason. However, this method suffers from both substantial computational resource computing and catastrophic forgetting. The fundamental cause is that, with the primary goal of enhancing multilingual comprehension, an excessive number of irrelevant layers and parameters are tuned during the first stage. Given our findings that the representation learning of languages is merely conducted in lower-level layers, we propose an efficient multilingual reasoning alignment approach that precisely identifies and fine-tunes the layers responsible for handling multilingualism. Experimental results show that our method, SLAM, only tunes 6 layers' feed-forward sub-layers including 6.5−8% of all parameters within 7B and 13B LLMs, achieving superior average performance than all strong baselines across 10 languages. Meanwhile, SLAM only involves one training stage, reducing training time by 4.1 − 11.9× compared to the two-stage method. © 2025 Association for Computational Linguistics.

关键词： Computational linguistics

来源：评论

学校读者我要写书评

暂无评论

Improved Regret for Bandit Convex Optimization with Delayed Feedback 38

Improved Regret for Bandit Convex Optimization with Delayed ...

引用

38th Conference on Neural Information Processing Systems, NeurIPS 2024

作者： Wan, Yuanyu Yao, Chang Song, Mingli Zhang, Lijun School of Software Technology Zhejiang University Ningbo China State Key Laboratory of Blockchain and Data Security Zhejiang University Hangzhou China Institute of Blockchain and Data Security Hangzhou China National Key Laboratory for Novel Software Technology Nanjing University Nanjing China

We investigate bandit convex optimization (BCO) with delayed feedback, where only the loss value of the action is revealed under an arbitrary delay. Let n, T, d¯ denote the dimensionality, time horizon, and average delay, respectively. Previous studies have achieved an O(√nT3/4 +(nd¯)1/3T2/3) regret bound for this problem, whose delay-independent part matches the regret of the classical non-delayed bandit gradient descent algorithm. However, there is a large gap between its delay-dependent part, i.e., O((nd¯)1/3T2/3), and an existing Ω(√d¯T) lower bound. In this paper, we illustrate that this gap can be filled in the worst case, where d¯is very close to the maximum delay d. Specifically, we first develop a novel algorithm, and prove that it enjoys a regret bound of O(√nT3/4 + √dT) in general. Compared with the previous result, our regret bound is better for d = O((nd¯)2/3T1/3), and the delay-dependent part is tight in the worst case. The primary idea is to decouple the joint effect of the delays and the bandit feedback on the regret by carefully incorporating the delayed bandit feedback with a blocking update mechanism. Furthermore, we show that the proposed algorithm can improve the regret bound to O((nT)2/3 log1/3 T + dlog T) for strongly convex functions. Finally, if the action sets are unconstrained, we demonstrate that it can be simply extended to achieve an O(n√T log T + dlog T) regret bound for strongly convex and smooth functions. © 2024 Neural information processing systems foundation. All rights reserved.

关键词：

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：