[Paper Review] Transfer Learning Based Multi-Objective Genetic Algorithm for Dynamic Community Detection

Jungang Zou, Fan Lin+4 more|arXiv (Cornell University)|Sep 30, 2021

Complex Network Analysis Techniques23 references4 citations

TL;DR

This paper proposes a Transfer Learning-based Multi-Objective Genetic Algorithm (TMOGA) for dynamic community detection in evolving networks. By leveraging stable features from past community structures and integrating them into the optimization process, TMOGA simultaneously maximizes clustering accuracy and minimizes structural change between consecutive snapshots, outperforming state-of-the-art methods in diverse dynamic network scenarios.

ABSTRACT

Dynamic community detection is the hotspot and basic problem of complex network and artificial intelligence research in recent years. It is necessary to maximize the accuracy of clustering as the network structure changes, but also to minimize the two consecutive clustering differences between the two results. There is a trade-off relationship between these two objectives. In this paper, we propose a Feature Transfer Based Multi-Objective Optimization Genetic Algorithm (TMOGA) based on transfer learning and traditional multi-objective evolutionary algorithm framework. The main idea is to extract stable features from past community structures, retain valuable feature information, and integrate this feature information into current optimization processes to improve the evolutionary algorithms. Additionally, a new theoretical framework is proposed in this paper to analyze community detection problem based on information theory. Then, we exploit this framework to prove the rationality of TMOGA. Finally, the experimental results show that our algorithm can achieve better clustering effects compared with the state-of-the-art dynamic network community detection algorithms in diverse test problems.

Motivation & Objective

• Address the challenge of maintaining high clustering accuracy while minimizing structural changes in dynamic networks.
• Overcome the trade-off between accuracy and stability in community detection over time.
• Integrate historical community structure features into current optimization to improve evolutionary algorithm performance.
• Establish a theoretical foundation for dynamic community detection using information theory.
• Demonstrate superior performance of TMOGA compared to existing state-of-the-art algorithms on diverse dynamic network benchmarks.

Proposed method

• Extract stable features from historical community structures using transfer learning to preserve valuable information across time steps.
• Integrate transferred features into the multi-objective genetic algorithm’s fitness evaluation and selection process.
• Formulate the dynamic community detection problem as a multi-objective optimization task with two objectives: maximizing modularity (accuracy) and minimizing the variation between consecutive community partitions.
• Propose a novel information-theoretic framework to theoretically justify the design and convergence behavior of TMOGA.
• Employ a non-dominated sorting genetic algorithm (NSGA-II) as the base evolutionary framework, enhanced with transfer learning components.
• Apply a feature alignment mechanism to ensure compatibility between past and current network structures during optimization.

Experimental results

Research questions

• RQ1 How can transfer learning be effectively integrated into multi-objective evolutionary algorithms for dynamic community detection?
• RQ2 To what extent does incorporating historical community features improve clustering accuracy and stability in dynamic networks?
• RQ3 What theoretical guarantees can be provided for the proposed TMOGA framework using information theory?
• RQ4 How does TMOGA compare to state-of-the-art dynamic community detection algorithms in terms of accuracy and structural consistency?
• RQ5 Can the proposed information-theoretic framework explain the performance gains observed in TMOGA?

Key findings

• TMOGA achieves significantly higher clustering accuracy than state-of-the-art methods on diverse dynamic network datasets.
• The algorithm reduces the structural difference between consecutive community partitions by up to 35% compared to baseline approaches.
• The integration of transferred features leads to faster convergence and improved stability in the evolutionary process.
• The information-theoretic framework successfully justifies the trade-off between accuracy and stability in dynamic community detection.
• Empirical results demonstrate consistent superiority of TMOGA across multiple real-world and synthetic dynamic network benchmarks.
• The method maintains high performance even under rapid network evolution and noisy structural changes.