Graph reducing method based on deep reinforcement learning
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
The development wave of general artificial intelligence drives the generation and processing of massive data, and large-scale and heterogeneous graph data networks constitute an important foundation of the digital world. However, the continuously growing scale of data not only increases the difficulty of graph data processing, but also creates the need to reduce graph size and maximize the amount of graph information. Existing methods make it difficult to synergistically control the graph size and optimize the amount of graph information, which limits the effectiveness of graph data analysis and processing. In response to the need for balanced control of graph data scale and information content, the graph reducing problem with scale regulation as the constraint and information maximization as the goal was proposed. Specifically, a graph fusion algorithm and a deep reinforcement learning-based graph reducing algorithm were designed to solve the problem, including graph reducing operations such as node fusion, composite mapping, and methods used for similarity metrics. Experiments verified the balanced regulation ability of the reducing algorithm, and comparisons with four algorithms across three evaluation metrics—feature similarity, graph similarity, and edge information loss—showed that the proposed graph reduction method could achieve performance improvements of at least 20.7%, 19.9%, and 26.3%, respectively.
Fields of Study & Subjects
Graph Theory and Algorithms
Advanced Graph Neural Networks
Complex Network Analysis Techniques
Publication details
- Journal / source
- DOAJ (DOAJ: Directory of Open Access Journals)
- Publication date
- 1 March 2026
- Language
- ZH