A Comprehensive Survey on Community Detection With Deep Learning.

Area of Science:

• Network analysis and machine learning
• Artificial intelligence applications in data science

Background:

• Community detection is crucial for understanding network structures and member behaviors.
• Traditional methods like spectral clustering face challenges with high-dimensional network data.
• Deep learning offers advanced solutions for complex community detection tasks.

Purpose of the Study:

• To provide a comprehensive review of recent advancements in deep learning for network community detection.
• To introduce a novel taxonomy for classifying state-of-the-art deep learning methods.
• To summarize experimental settings, applications, and future research directions.

Main Methods:

• Categorization of deep learning models including deep neural networks (DNNs), deep nonnegative matrix factorization, and deep sparse filtering.
• Further classification of DNNs into convolutional networks, graph attention networks, generative adversarial networks, and autoencoders.
• Summary of benchmark datasets, evaluation metrics, and open-source implementations.

Main Results:

• A structured overview of deep learning techniques applied to community detection.
• Identification of key methodologies and their variations within the deep learning paradigm.
• Compilation of resources for experimental reproducibility and comparative analysis.

Conclusions:

• Deep learning methods represent a significant leap forward in network community detection, particularly for complex datasets.
• The proposed taxonomy aids in navigating the diverse landscape of deep learning approaches.
• Future research should focus on emerging challenges and opportunities in this dynamic field.