检索结果-内蒙古大学图书馆

GPS: graph contrastive learning via multi-scale augmented views from adversarial pooling

Science China(Information Sciences) 2025年第1期68卷 145-158页

作者： Wei JU Yiyang GU Zhengyang MAO Ziyue QIAO Yifang QIN Xiao LUO Hui XIONG Ming ZHANG School of Computer Science National Key Laboratory for Multimedia Information ProcessingPeking University Artificial Intelligence Thrust The Hong Kong University of Science and Technology Department of Computer Science University of California

Self-supervised graph representation learning has recently shown considerable promise in a range of fields, including bioinformatics and social networks. A large number of graph contrastive learning approaches have shown promising performance for representation learning on graphs, which train models by maximizing agreement between original graphs and their augmented views(i.e., positive views). Unfortunately, these methods usually involve pre-defined augmentation strategies based on the knowledge of human experts. Moreover, these strategies may fail to generate challenging positive views to provide sufficient supervision signals. In this paper, we present a novel approach named graph pooling contrast(GPS) to address these *** by the fact that graph pooling can adaptively coarsen the graph with the removal of redundancy, we rethink graph pooling and leverage it to automatically generate multi-scale positive views with varying emphasis on providing challenging positives and preserving semantics, i.e., strongly-augmented view and weakly-augmented view. Then, we incorporate both views into a joint contrastive learning framework with similarity learning and consistency learning, where our pooling module is adversarially trained with respect to the encoder for adversarial robustness. Experiments on twelve datasets on both graph classification and transfer learning tasks verify the superiority of the proposed method over its counterparts.

关键词： graph representation learning graph neural networks graph contrastive learning graph augmentations graph pooling

来源：评论

学校读者我要写书评

暂无评论

Semi-Supervised Dual-Stream Self-Attentive Adversarial graph contrastive learning for Cross-Subject EEG-Based Emotion Recognition

引用

IEEE TRANSACTIONS ON AFFECTIVE COMPUTING 2025年第1期16卷 290-305页

作者： Ye, Weishan Zhang, Zhiguo Teng, Fei Zhang, Min Wang, Jianhong Ni, Dong Li, Fali Xu, Peng Liang, Zhen Shenzhen Univ Sch Biomed Engn Med Sch Shenzhen 518060 Peoples R China Guangdong Prov Key Lab Biomed Measurements & Ultra Shenzhen 518060 Peoples R China Harbin Inst Technol Inst Comp & Intelligence Shenzhen 518000 Peoples R China Marshall Lab Biomed Engn Shenzhen 518060 Peoples R China Peng Cheng Lab Shenzhen 518055 Peoples R China Shenzhen Kangning Hosp Shenzhen Mental Hlth Ctr Shenzhen 518020 Peoples R China Univ Elect Sci & Technol China Clin Hosp Chengdu Brain Sci Inst MOE Key Lab Neuroinformat Chengdu 611731 Peoples R China Univ Elect Sci & Technol China Sch Life Sci & Technol Ctr Informat Med Chengdu 611731 Peoples R China

Electroencephalography (EEG) is an objective tool for emotion recognition with promising applications. However, the scarcity of labeled data remains a major challenge in this field, limiting the widespread use of EEG-based emotion recognition. In this paper, a semi-supervised Dual-stream Self-attentive Adversarial graph contrastive learning framework (termed as DS-AGC) is proposed to tackle the challenge of limited labeled data in cross-subject EEG-based emotion recognition. The DS-AGC framework includes two parallel streams for extracting non-structural and structural EEG features. The non-structural stream incorporates a semi-supervised multi-domain adaptation method to alleviate distribution discrepancy among labeled source domain, unlabeled source domain, and unknown target domain. The structural stream develops a graph contrastive learning method to extract effective graph-based feature representation from multiple EEG channels in a semi-supervised manner. Further, a self-attentive fusion module is developed for feature fusion, sample selection, and emotion recognition, which highlights EEG features more relevant to emotions and data samples in the labeled source domain that are closer to the target domain. Extensive experiments are conducted on four benchmark databases (SEED, SEED-IV, SEED-V, and FACED) using a semi-supervised cross-subject leave-one-subject-out cross-validation evaluation protocol. The results show that the proposed model outperforms existing methods under different incomplete label conditions with an average improvement of 2.17%, which demonstrates its effectiveness in addressing the label scarcity problem in cross-subject EEG-based emotion recognition.

关键词： Feature extraction Emotion recognition Brain modeling Electroencephalography Streams Data models Transfer learning EEG emotion recognition graph contrastive learning domain adaption semi-supervised learning

来源：评论

学校读者我要写书评

暂无评论

Self-attentive Rationalization for Interpretable graph contrastive learning

引用

ACM TRANSACTIONS ON KNOWLEDGE DISCOVERY FROM DATA 2025年第2期19卷 1-21页

作者： Li, Sihang Luo, Yanchen Zhang, An Wang, Xiang Li, Longfei Zhou, Jun Chua, Tat-seng Univ Sci & Technol China Hefei Peoples R China Natl Univ Singapore Singapore Singapore Ant Grp Hangzhou Peoples R China

graph augmentation is the key component to reveal instance-discriminative features of a graph as its rationale- an interpretation for it-in graph contrastive learning (GCL). Existing rationale-aware augmentation mechanisms in GCL frameworks roughly fall into two categories and suffer from inherent limitations: (1) non-heuristic methods with the guidance of domain knowledge to preserve salient features, which require expensive expertise and lack generality, or (2) heuristic augmentations with a co-trained auxiliary model to identify crucial substructures, which face not only the dilemma between system complexity and transformation diversitybut also the instability stemming from the co-training of two separated sub-models. Inspired by recent studies on transformers, we propose self-attentive rationale-guided GCL (SR-GCL), which integrates rationale generator and encoder together, leverages the self-attention values in transformer module as a natural guidance to delineate semantically informative substructures from both node- and edge-wise perspectives, and contrasts on rationale-aware augmented pairs. On real-world biochemistry datasets, visualization results verify the effectiveness and interpretability of self-attentive rationalization, and the performance on downstream tasks demonstrates the state-of-the-art performance of SR-GCL for graph model pre-training. Codes are available at https://***/lsh0520/SR-GCL.

关键词： Self-supervised learning interpretability graph contrastive learning self-attention mechanism

来源：评论

学校读者我要写书评

暂无评论

Molecular graph contrastive learning with line graph

引用

PATTERN RECOGNITION 2025年 162卷

作者： Chen, Xueyuan Li, Shangzhe Liu, Ruomei Shi, Bowen Liu, Jiaheng Wu, Junran Xu, Ke Beihang Univ State Key Lab Complex & Crit Software Environm 37 Xueyuan Rd Beijing 100191 Peoples R China Cent Univ Finance & Econ Sch Stat & Math 39 South Coll Rd Beijing 100081 Peoples R China Commun Univ China Sch Journalism 1 Dingfuzhuang East St Beijing 100024 Peoples R China

Trapped by the label scarcity in molecular property prediction and drug design, graph contrastive learning (GCL) came forward. Leading contrastive learning works show two kinds of view generators, that is, random or learnable data corruption and domain knowledge incorporation. While effective, the two ways also lead to molecular semantics altering and limited generalization capability, respectively. To this end, we relate the LinE graph with MOlecular graph contrastive learning and propose a novel method termed LEMON. Specifically, by contrasting the given graph with the corresponding line graph, the graph encoder can freely encode the molecular semantics without omission. Furthermore, we present anew patch with edge attribute fusion and two local contrastive losses enhance information transmission and tackle hard negative samples. Compared with state-of-the-art (SOTA) methods for view generation, superior performance on molecular property prediction suggests the effectiveness of our proposed framework.

关键词： Molecular pre-training graph contrastive learning Dual-helix graph encoder Hard negative samples Transfer learning

来源：评论

学校读者我要写书评

暂无评论

Attributed network community detection based on graph contrastive learning and multi-objective evolutionary algorithm

引用

NEUROCOMPUTING 2025年 636卷

作者： Liang, Yao Shu, Jian Liu, Linlan Nanchang Hangkong Univ Dept Software Nanchang 330063 Peoples R China Nanchang Hangkong Univ Dept Informat Engn Nanchang 330063 Peoples R China

Attributed network community detection holds significant research value for network structure analysis and practical applications. However, existing methods still face significant challenges in addressing the conflicts between topological structure and attribute features, as well as balancing structural tightness and attribute similarity in community detection. In light of this, we propose a community detection method based on graph contrastive learning and multi-objective evolutionary algorithm (GCL-MOEA) for attributed networks. Specifically, GCL-MOEA contains two core parts: node embedding and community detection. Considering the conflict between topological structure and attribute features, the node embedding part constructs topology-augmented and attribute-augmented views, which are utilized in a cross-view graph contrastive learning model. This model comprehensively extracts node features to obtain node embedding vectors, effectively preserving the consistency and complementarity between the structure and attributes. The community detection part utilizes clustering results of node embeddings to construct high-quality initial populations. A multi-objective evolutionary algorithm is subsequently employed to obtain community structures where nodes are tightly connected and have similar attributes. The effectiveness of the proposed method is validated on five real-world networks. Experimental results demonstrate that GCL-MOEA outperforms baselines in terms of ACC, NMI, ARI, and F1, obtaining better community detection results.

关键词： Attributed network Community detection graph contrastive learning Multi-objective evolutionary

来源：评论

学校读者我要写书评

暂无评论

Drug-drug interaction prediction based on graph contrastive learning and dual-view fusion

引用

COMPUTATIONAL BIOLOGY AND CHEMISTRY 2025年 117卷 108426页

作者： Ding, Shanyang Niu, Dongjiang Li, Mingxuan Zhang, Zhixin Li, Zhen Qingdao Univ Coll Comp Sci & Technol 308 Ningxia Rd Qingdao 266071 Shandong Peoples R China

Drug-drug interaction (DDI) is important in drug research and are one of the major causes of morbidity and mortality. The deep learning methods can automatically extract drug features from molecular graphs or drug-related networks, which improves the performance of DDI prediction. However, there is noise and incomplete data in existing datasets, and the volume of dataset is limited. In order to fully utilize the knowledge graph network and the molecular structure, we propose a dual-view fusion model GDF-DDI. In one view, the knowledge graph network and drug similarity network are constructed as the global information, and two graph convolution operations are implemented on both networks to extract drug embeddings. Subsequently, layer wise graph contrastive learning is performed to update the drug embeddings to captures richer semantic information. In the other view, the self-supervised learning is utilized to extract more comprehensive embedding of drugs. The embeddings under two views are concatenated to cover the global and local DDI information. The comparative experiments on two datasets show that our model outperforms other recent and state-of-the-art baselines.

关键词： Knowledge graph Drug similarity network graph contrastive learning Self-supervised learning Dual-view fusion model

来源：评论

学校读者我要写书评

暂无评论

Two-dimensional adversarial domain adaptation graph contrastive learning for fault diagnosis of limited similar batch process

引用

PROCESS SAFETY AND ENVIRONMENTAL PROTECTION 2025年 197卷

作者： Gao, Xingke Zhu, Jinlin Gao, Furong Zhang, Zheng Jiangnan Univ Sch Artificial Intelligence & Comp Sci Wuxi Peoples R China Jiangnan Univ Sch Food Sci & Technol State Key Lab Food Sci & Resources Wuxi Peoples R China Hong Kong Univ Sci & Technol Dept Chem & Biol Engn Hong Kong Peoples R China

Batch process systems can encounter various faults that affect safety differently. Identifying the types of faults is crucial, while new processes often lack sufficient labels for differentiation. To mitigate label scarcity, migrating labels from old, well-labeled processes appears to be a cost-effective solution. In this regard, this paper proposes a novel fault diagnosis approach based on graph contrastive learning, termed the Two-Dimensional Adversarial Domain Adaptation-based graph contrastive learning Network (2D-ADGCL). First, a 2D sliding window is employed to capture nonlinear dynamics between samples, and Statistical Pattern Analysis (SPA) is used to construct new samples containing more comprehensive statistical information. Second, a multi-head attention mechanism enhances the views of the graph structure. This facilitates the construction of positive and negative sample pairs for each instance, thus enabling the model to learn broader feature diversity and improving classification performance. Third, adversarial domain adaptation techniques are utilized to learn domain-invariant features between the source and target domains, enabling the classification of target domain samples for fault diagnosis. Finally, the sensitivity and superiority of 2D-ADGCL are validated through results from two experiments.

关键词： Fault diagnosis Similar batch process Two-dimensional modeling Adversarial domain adaptation Transfer leaning graph contrastive learning

来源：评论

学校读者我要写书评

暂无评论

graph contrastive learning Method with Sample Disparity Constraint and Feature Structure graph for Node Classification 16th

Graph Contrastive Learning Method with Sample Disparity Cons...

引用

16th International Conference on Knowledge Science, Engineering and Management (KSEM)

作者： Chen, Gangbin Cheng, Junwei Liang, Wanying He, Chaobo Tang, Yong South China Normal Univ Sch Comp Sci Guangzhou 510631 Peoples R China Pazhou Lab Guangzhou 510330 Guangdong Peoples R China

ISBN: (纸本)9783031402913;9783031402920

Most of the existing graph contrastive learning methods for node classification focus on exploiting topological information of the attributed networks, with little attention to the attribute information of the networks. Furthermore, in most graph contrastive learning methods, both positive and negative sample representations are generated by the same encoder, which makes them somewhat coupled and not conducive to distinguishing between them. But to our knowledge, most of these methods ignore the coupling between positive and negative sample representations, which may negatively affect the performance of the methods. To address these two issues, we propose a new method called DCFSG, in which we generate a feature structure graph from the attribute matrix, and then learn the node representations jointly from the topological structure graph and the feature structure graph. In addition, we impose a disparity constraint on the positive and negative samples and add corresponding loss to the objective function, which can help to lower the coupling between positive and negative samples and improves the performance of our method. The experimental results on real networks validate the effectiveness of our method.

关键词： graph contrastive learning Node classification Attributed networks Feature structure graph Disparity constraint

来源：评论

学校读者我要写书评

暂无评论

Semantic-Enhanced graph contrastive learning With Adaptive Denoising for Drug Repositioning

引用

IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS 2025年第3期29卷 1635-1643页

作者： Yu, Huimin Lu, Mingyu Li, Zeqian Zhang, Yijia Dalian Maritime Univ Sch Informat Sci & Technol Dalian 116026 Peoples R China Dalian Maritime Univ Sch Artificial Intelligence Dalian 116024 Peoples R China

The traditional drug development process requires a significant investment in workforce and financial resources. Drug repositioning as an efficient alternative has attracted much attention during the last few years. Despite the wide application and success of the method, there are still many shortcomings in the existing model. For example, sparse datasets will seriously affect the existing methods' performance. Additionally, these methods do not pay attention to the noise in datasets. In response to the above defects, we propose a semantic-enriched augmented graph contrastive learning with an adaptive denoising method, called SGCD. This method enhances data from the perspective of the embedding layer, deeply mines potential neighborhood relation-ships in semantic space, and combines similar drugs in the semantic neighborhoods into prototype comparison targets, thus effectively mitigating the impact of data sparsity on the model. Moreover, to enhance the model's robustness to noisy data, we use the adaptive denoising method, which can effectively identify noisy data in the training process. Exhaustive experiments on multiple real datasets show the effectiveness of the proposed model.

关键词： Drugs Semantics Diseases Data models Noise measurement Biological system modeling Bioinformatics Drug repositioning graph contrastive learning semantic learning

来源：评论

学校读者我要写书评

暂无评论

Understanding and mitigating dimensional collapse of graph contrastive learning: A non-maximum removal approach

引用

NEURAL NETWORKS 2025年 181卷 106652页

作者： Sun, Jiawei Chen, Ruoxin Li, Jie Ding, Yue Wu, Chentao Liu, Zhi Yan, Junchi Shanghai Jiao Tong Univ Dept Comp Sci & Engn Shanghai Peoples R China Univ Electrocommun Dept Comp & Network Engn Tokyo Japan

graph contrastive learning (GCL) generates graph-level embeddings by maximizing Mutual Information between different augmented views of the same graph (positive pairs), and shows promising performance in graph representation learning (GRL) without the supervision of manual annotations. However, GCL suffers from a dimensional collapse problem, i.e., embedding vectors reside in a restricted low-dimensional subspace, curtailing the expressiveness of certain embedding dimensions. In this paper, we present a theoretical analysis identifying the smoothing effect of graph pooling and graph convolution's implicit regularization as principal causes of dimension collapse in graph contrastive learning. To mitigate the above effects, we propose a Non-Maximum Removal graph contrastive learning (nmrGCL) approach, which removes "prominent"dimensions " dimensions (those significantly contributing to the similarity measure) for positive pairs within the pretext task. Comprehensive experiments on multiple benchmark datasets are conducted and the results show that the proposed nmrGCL outperforms the state-of-the-art methods.

关键词： graph Neural Networks graph contrastive learning graph Representation learning Dimensional collapse

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：