Scholarlink

<p>Node2Vec, an advanced graph analysis technique rooted in graph-based learning principles, excels at node feature extraction. However, it faces significant challenges, particularly in parameter optimization and maintaining efficiency in large-scale networks. To address these issues, this paper proposes a novel integration of Node2Vec with the Grey Wolf Optimization algorithm. This hybrid approach systematically explores the parameter space to identify optimal configurations for Node2Vec, enhancing the quality of embeddings for diverse graph and network analyses. Experimental results demonstrate that the proposed method significantly improves Node2Vec's performance, outperforming the original implementation, grid search variants, and other leading algorithms. Comprehensive evaluations on benchmark datasets, including both synthetic and real-world networks, confirm that our approach achieves superior performance metrics compared to existing solutions.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A hybrid optimization approach for graph embedding: leveraging Node2Vec and grey wolf optimization

  • 期刊论文

  • 发表日期：

    2025年2月17日

查看全文
  • Mahdi Rabiei,
  • Mehdi Fartash,
  • Sara Nazari

摘要

Node2Vec, an advanced graph analysis technique rooted in graph-based learning principles, excels at node feature extraction. However, it faces significant challenges, particularly in parameter optimization and maintaining efficiency in large-scale networks. To address these issues, this paper proposes a novel integration of Node2Vec with the Grey Wolf Optimization algorithm. This hybrid approach systematically explores the parameter space to identify optimal configurations for Node2Vec, enhancing the quality of embeddings for diverse graph and network analyses. Experimental results demonstrate that the proposed method significantly improves Node2Vec's performance, outperforming the original implementation, grid search variants, and other leading algorithms. Comprehensive evaluations on benchmark datasets, including both synthetic and real-world networks, confirm that our approach achieves superior performance metrics compared to existing solutions.