Sampling clustering based on multi-view attribute structural relations.

Area of Science:

• Data Science
• Machine Learning
• Graph Analytics

Background:

• Graph data complexity is rising exponentially.
• Multi-view graph clustering is crucial for real-world data with diverse relationships.
• Existing deep learning methods struggle with multi-view graph data, lacking joint structure-attribute modeling and varied feature handling.

Purpose of the Study:

• To propose a novel, effective, and straightforward multi-view graph clustering approach.
• To address the limitations of current deep learning techniques in handling multi-view graph data.
• To improve the accuracy and efficiency of graph clustering.

Main Methods:

• The proposed SLMGC approach utilizes graph filtering to reduce noise.
• It employs node importance sampling to decrease computational complexity.
• Graph contrastive regularization enhances clustering representations, and a self-training algorithm achieves final clustering.

Main Results:

• SLMGC demonstrates superior performance in multi-view graph clustering tasks.
• The method effectively handles multi-view graph data with varying features.
• Experimental results validate the supremacy of SLMGC over prevailing deep neural network techniques.

Conclusions:

• SLMGC offers a robust and efficient solution for multi-view graph clustering.
• The approach simplifies parameter settings compared to deep learning alternatives.
• SLMGC represents a significant advancement in processing complex, multi-view graph data.