Article
Computer Science, Artificial Intelligence
Hui Zhou, Maoguo Gong, Shanfeng Wang, Yuan Gao, Zhongying Zhao
Summary: Graph contrastive learning (GCL) aims to generate supervision information by transforming graph data itself, and it has become a focus of graph research recently. However, most GCL methods are unsupervised and struggle with balancing multi-view graph information. To address this, we propose a semi-supervised multi-view graph contrastive learning (SMGCL) framework for graph classification. The framework captures comparative relations between label-independent and label-dependent node pairs across different views and incorporates a label augmentation module and a shared decoder module to enhance discriminative representations and extract underlying relationships between representations and graph topology. Experimental results demonstrate the superiority of our proposed framework for graph classification tasks.
KNOWLEDGE-BASED SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Guangfeng Lin, Xiaobing Kang, Kaiyang Liao, Fan Zhao, Yajun Chen
Summary: Graph learning dynamically captures data distribution structure based on graph convolutional networks. The quality of learning the graph structure directly impacts semi-supervised classification using GCN. Existing methods combine computational layers and losses into GCN to explore global and local graphs, which have different roles in semi-supervised classification.
PATTERN RECOGNITION
(2021)
Article
Computer Science, Information Systems
Yu Xie, Shengze Lv, Yuhua Qian, Chao Wen, Jiye Liang
Summary: Researchers propose a novel active and semi-supervised graph neural network framework, which can effectively perform graph classification tasks using a small number of labeled and unlabeled graph examples, and achieves competitive performance on benchmark graph datasets.
IEEE TRANSACTIONS ON BIG DATA
(2022)
Article
Computer Science, Information Systems
Xiuyi Jia, Tao Wen, Weiping Ding, Huaxiong Li, Weiwei Li
Summary: Label distribution learning (LDL) is a new paradigm in machine learning that addresses label ambiguity by emphasizing the relevance of each label to a particular instance. We propose a projection graph embedding algorithm for semi-supervised label distribution learning (PGE-SLDL), which aims to select valuable features, construct an accurate graph, and recover unknown label distributions. The self-updating projection graph is more effective in learning label distribution compared to traditional fixed graphs in semi-supervised learning.
INFORMATION SCIENCES
(2021)
Article
Computer Science, Artificial Intelligence
Wei Ju, Xiao Luo, Zeyu Ma, Junwei Yang, Minghua Deng, Ming Zhang
Summary: This paper proposes a Graph Harmonic Neural Network (GHNN) that combines the advantages of graph convolutional networks and graph kernels to fully utilize unlabeled data, overcoming the scarcity of labeled data in semi-supervised scenarios.
Review
Computer Science, Artificial Intelligence
Zixing Song, Xiangli Yang, Zenglin Xu, Irwin King
Summary: This article introduces a graph-based semi-supervised learning (GSSL) method, which represents each sample as a node in a graph and infers the label information of unlabeled samples based on the graph's structure. The article provides an in-depth understanding of GSSL methods and their advancements, as well as insights into future research directions in this field.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Bo Jiang, Si Chen, Beibei Wang, Bin Luo
Summary: The problem of multiple graph learning involves learning consistent representation by exploiting the complementary information of multiple graphs. This paper proposes a novel learning framework, called Multiple Graph Learning Neural Networks (MGLNN), which aims to learn an optimal graph structure from multiple graph structures and integrate multiple graph learning and Graph Neural Networks' representation. Experimental results demonstrate that MGLNN outperforms other methods on semi-supervised classification tasks.
Article
Biochemical Research Methods
Joung Min Choi, Chaelin Park, Heejoon Chae
Summary: This paper presents meth-SemiCancer, a semi-supervised cancer subtype classification method based on DNA methylation profiles. The proposed model is pre-trained on methylation datasets with cancer subtype labels, generates pseudo-subtypes for datasets without subtype information, and performs fine-tuning using both labeled and unlabeled datasets. Compared to other methods, meth-SemiCancer achieves higher F1-score and Matthews correlation coefficient.
BMC BIOINFORMATICS
(2023)
Article
Computer Science, Artificial Intelligence
Ruigang Zheng, Weifu Chen, Guocan Feng
Summary: Inspections on current graph neural networks suggest re-evaluating the computational aspect of final aggregation to focus on intra-class relations and produce smoother predictions. By incorporating metric learning and entropy losses, the proposed algorithm effectively reduces inter-class edge weights and improves classification accuracy.
PATTERN RECOGNITION
(2022)
Article
Computer Science, Artificial Intelligence
Guolin Zhang, Zehui Hu, Guoqiu Wen, Junbo Ma, Xiaofeng Zhu
Summary: The traditional graph convolutional network(GCN) and its variants usually only propagate node information through the given topology, ignoring some correlative feature information between nodes. To address this issue, a novel model named Dynamic Graph Convolutional Networks by Semi-Supervised Contrastive Learning (DGSCL) is proposed in this paper. It constructs a dynamic feature graph from the input node features and fuses it with the given topology using co-attention modules for more informative node embeddings.
PATTERN RECOGNITION
(2023)
Article
Automation & Control Systems
Xiaoyi Mai, Romain Couillet
Summary: A new regularization approach involving centering operation is proposed as a solution to the high-dimensional learning efficiency problem in semi-supervised learning, supported by both theoretical analysis and empirical results.
JOURNAL OF MACHINE LEARNING RESEARCH
(2021)
Article
Computer Science, Artificial Intelligence
Alexander Mey, Marco Loog
Summary: This survey explores semi-supervised learning and its applications in classification and regression tasks. It summarizes theoretical results and highlights the assumptions made when utilizing unlabeled data. The survey aims to identify the limits and potential benefits of semi-supervised learning, focusing on understanding the underlying theory and assumptions.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2023)
Article
Computer Science, Artificial Intelligence
Zheng Liu, Shiluo Huang, Wei Jin, Ying Mu
Summary: BLS faces challenges in handling datasets with few labeled samples, prompting the proposal of S2-BLS. The novel approach utilizes semi-supervised ELM-AE to obtain mapped features and calculates discriminative projecting weights between labeled instances and transformed features.
Article
Chemistry, Analytical
Yifan Wang, Yan Huang, Qicong Wang, Chong Zhao, Zhenchang Zhang, Jian Chen
Summary: This paper proposes a semi-supervised learning method to reduce noise in pseudo-labels. It improves the quality by considering the accuracy and confidence of predictions. It introduces a self-training framework and an uncertainty-based graph convolutional network to enhance the performance.
Article
Computer Science, Information Systems
Jingliu Lai, Hongmei Chen, Tianrui Li, Xiaoling Yang
Summary: In this study, a novel semi-supervised sparse feature selection framework is proposed, which improves the quality of the similarity matrix through adaptive graph learning and alleviates the negative influence of redundant features through redundancy minimization regularization.
INFORMATION SCIENCES
(2022)