Q-learning-based community detection algorithm.

Area of Science:

• Complex network analysis
• Machine learning
• Graph theory

Background:

• Community detection is crucial for understanding network structures.
• Conventional algorithms face challenges like local optima and high computational costs.
• Need for robust and scalable community detection methods.

Purpose of the Study:

• To develop an advanced community detection framework using multiagent reinforcement learning.
• To address limitations of existing methods, including initialization sensitivity and scalability.
• To improve the accuracy and efficiency of community detection in complex networks.

Main Methods:

• A multiagent reinforcement learning framework with independent agents for each community.
• Deep Q-learning network for adaptive node allocation.
• Reward function balancing intracommunity compactness and intercommunity separateness.
• Node importance-guided initialization and DeepWalk for node embeddings.
• Epsilon-greedy strategy and target network updates for enhanced exploration and stability.

Main Results:

• The proposed method consistently outperforms baseline approaches across multiple datasets.
• Achieved more accurate community detection compared to traditional algorithms.
• Demonstrated superior scalability, especially for large-scale networks.
• Effective handling of initialization sensitivity and local optima problems.

Conclusions:

• The multiagent reinforcement learning framework offers a significant advancement in community detection.
• The method provides a robust, accurate, and scalable solution for analyzing complex networks.
• This approach is particularly beneficial for large-scale network analysis where traditional methods falter.