Organizations: School of Computer, Jiangsu University of Science and Technology, Zhenjiang, 212100, China · Chongqing Key Laboratory of Computational Intelligence, Chongqing University of Posts and Telecommunications, Chongqing, 400065, China · College of Computer and Data Science, Fuzhou University, Fuzhou 350116, China
The effective utilization of consistency is crucial for multi-view learning. GCNs leverage node connections to propagate information across the graph, facilitating the exploitation of consistency in multi-view data. However, most existing GCN-based multi-view methods suffer from several limitations. First, current approaches predominantly rely on KNN for topology construction, where the artificial selection of the k value significantly constrains the effective exploitation of inter-node consistency. Second, the inter-feature consistency within individual views is often overlooked, which adversely affects the quality of the final embedding representations. Moreover, these methods fail to fully utilize inter-view consistency as the fusion of embedded representations from multiple views is often implemented after the intra-view graph convolutional operation. Collectively, these issues limit the model's capacity to fully capture inter-node, inter-feature and inter-view consistency. To address these issues, this paper proposes the multi-view graph convolutional network with fully leveraging consistency via GB-based topology construction, feature enhancement and interactive fusion (MGCN-FLC). MGCN-FLC can fully utilize three types of consistency via the following three modules to enhance learning ability:The topology construction module based on the granular ball algorithm, which clusters nodes into granular balls with high internal similarity to capture inter-node consistency;The feature enhancement module that improves feature representations by capturing inter-feature consistency;The interactive fusion module that enables each view to deeply interact with all other views, thereby obtaining more comprehensive inter-view consistency. Experimental results on nine datasets show that the proposed MGCN-FLC outperforms state-of-the-art semi-supervised node classification methods.
Figures & tables
Figure 1: MGCN-FLC consists of three modules: the topology construction module (TC), the feature enhancement module (FE) and the interactive fusion module (IF). Specifically, the TC module constructs the topology by establishing the inter-GB and intra-GB full connections between nodes, providing an effective pathway for information propagation in the GCN. The FE module applies mixed pooling to the stacking result of the similarity feature matrix and the original feature matrix, generating the enhanced feature representations. Subsequently, the enhanced feature representations generated by FE module from multiple views are encoding into the same dimension. Based on the encoder’s output, the IF module computes and aggregates interactive features between two distinct views to ultimately generate node embeddings. Finally, the topology constructed by the TC module and the node embeddings generated by the IF module are jointly fed as input into a standard GCN for node prediction.
Datasets
# samples
# views
# features
# classes
# data types
BBCnews
685
4
4,659/4,633/4,665/4,684
5
Textual documents
Caltech101-7
1474
6
48/40/254/1984/512/928
7
Object images
MNIST
10000
3
30/9/9
10
Digit images
NUS-WIDE
2400
6
64/73/128/144/225/500
8
Object images
WebKB
203
3
1,703/230/230
4
Textual documents
Table 1: A brief description of multi-view datasets
Methods/Datasets
BBCnews
Caltech101-7
MNIST
NUS-WIDE
WebKB
BBCSports
NGs
ProteinFold
Reuters
MLAN (2017)
74.1 73.5
76.3 75.6
87.7 86.1
34.1 33.2
73.9 41.2
80.5 80.1
85.5 83.3
35.6 31.2
64.6 63.2
MVAR (2017)
75.3 76.6
79.1 78.5
84.9 81.7
29.3 28.5
76.9 46.3
82.1 82.8
86.6 85.7
39.6 35.1
83.5 81.8
Co-GCN (2020)
81.9 78.1
83.1 80.2
90.2 90.1
27.5 28.1
72.6 39.3
86.3 84.9
84.2 83.6
41.2 37.8
78.4 73.7
DSRL (2022)
84.5 83.1
85.3 84.6
88.9 87.1
43.6 43.2
82.5 58.2
90.9 93.7
88.7 84.5
46.5 39.5
36.5 25.3
Table 2: The classification performance of all compared semi-supervised classification methods uses 10% labeled samples as supervision, where the best performance is highlighted in bold and the second-best results are underlined.
Figure 2: ACC of all methods as the ratio of labeled data ranges in {0.05, 0.10, …, 0.5} on the nine datasets.
Figure 3: F1-score of all methods as the ratio of labeled data ranges in {0.05, 0.10, …, 0.5} on the nine datasets.
Figure 4: The ACC and F1-score of MGCN-FLC w.r.t. hyperparameter α on datasets.
Figure 5: The ACC and F1-score of MGCN-FLC w.r.t. hyperparameter β on datasets.
Figure 6: The curves of MGCN-FLC’s loss values, ACC and F1-score on the selected datasets.
Figure 7: t-SNE visualizations of the features from JFGCN, LGCN-FF, and MGCN-FLC on the (a-c) BBCnews and (d-f) NUS-WIDE datasets, with colors denoting different classes.
Datasets/Methods
MGCN-FLCtc
MGCN-FLCfe
MGCN-FLCif
MGCN-FLC
BBCnews ACC F1
91.1 90.2
95.3 94.3
98.4 97.9
98.4 97.9
Caltech101-7 ACC F1
83.8 73.1
92.0 83.3
97.4 90.7
97.5 91.3
MNIST ACC F1
90.6 88.6
95.0 94.3
82.9 76.9
96.4 95.8
NUS-WIDE ACC F1
40.9 38.9
98.1 98.1
98.1 98.1
99.0 99.0
WebKB ACC F1
82.0 44.6
86.9 66.3
91.8 67.4
91.8 67.4
Table 3: Ablation study results of the proposed MGCN-FLC on all datasets.
Figure 8: The homophily ratios of the topologies generated by the four algorithms—JFGCN, GBCM-GCN, LGCN-FF, and MGCN-FLC—across the WebKB, BBCnews, and Caltech101-7 datasets.
Datasets/Methods
MGCN-FLC
center node
density peak node
representative node (our)
BBCnews ACC F1
98.4 97.9
98.4 97.9
98.4 97.9
Caltech101-7 ACC F1
97.5 91.3
97.5 91.3
97.5 91.3
MNIST ACC F1
95.4 94.8
95.9 94.9
96.4 95.8
NUS-WIDE ACC F1
99.0 99.0
98.5 98.5
99.0 99.0
Table 4: The classification results of MGCN-FLC based on representative node and its variants (center node and density peak node) on all datasets.
Datasets/Methods
MGCN-FLC
VA-based
VSF-based
IF-based (our)
BBCnews ACC F1
98.0 97.2
96.6 95.5
98.4 97.9
Caltech101-7 ACC F1
79.0 55.2
69.8 40.2
97.5 91.3
MNIST ACC F1
95.6 93.5
81.2 78.5
96.4 95.8
NUS-WIDE ACC F1
79.0 78.9
77.6 77.6
99.0 99.0
Table 5: The classification results of MGCN-FLC and its variants on all datasets. ’VA-based’ denotes the MGCN-FLC variant based on view-level attention. ’VSF-based’ denotes the MGCN-FLC variant based on view-shared underlying features. ’IF-based’ denotes the original MGCN-FLC model.
Figure 9: Running time of MGCN-FLC and other GCN-based models on all datasets.
JiangSu University of Science and Technoloy, Zhenjiang, Jiangsu, China · Chongqing University of Posts and Telecommunications, Chongqing, China · Zunyi Normal University, Zunyi, Guizhou, China
School of Computer and Cyber Security, Fujian Normal University, Fuzhou 350000, China · Digital Fujian Internet-of-Things Laboratory of Environmental Monitoring, Fujian Normal University, Fuzhou 350000, China · School of Computer Science, McGill University, Montreal H3A 2A7, Quebec, Canada