Adaptive robust structure exploration for complex systems based on model configuration and fusion

Yingfei Qu; Wanbing Liu; Junhao Wen; Ming Li

doi:10.7717/peerj-cs.1983

Adaptive robust structure exploration for complex systems based on model configuration and fusion

PeerJ Comput Sci. 2024 Apr 8:10:e1983. doi: 10.7717/peerj-cs.1983. eCollection 2024.

Authors

Yingfei Qu¹, Wanbing Liu², Junhao Wen¹, Ming Li³

Affiliations

¹ Computer Science and Technology Post-Doctoral Station, Chongqing University, Chongqing, China.
² Hengda Fuji Elevator Co. Ltd., Huzhou, China.
³ Chongqing Key Laboratory for Intelligent Perception and Blockchain Technology, Chongqing Technology and Business University, Chongqing, China.

Abstract

Analyzing and obtaining useful information is challenging when facing a new complex system. Traditional methods often focus on specific structural aspects, such as communities, which may overlook the important features and result in biased conclusions. To address this, this article suggests an adaptive algorithm for exploring complex system structures using a generative model. This method calculates and optimizes node parameters, which can reflect the latent structural characteristics of the complex system. The effectiveness and stability of this method have been demonstrated in comparative experiments on 10 sets of benchmark networks using our model parameter configuration scheme. To enhance adaptability, algorithm fusion strategies were also proposed and tested on two real-world networks. The results indicate that the algorithm can uncover multiple structural features, including clustering, overlapping, and local chaining. This adaptive algorithm provides a promising approach for exploring complex system structures.

Keywords: Algorithm fusion; Complex network; Complex system; Model configuration; Multiple structural features.

Grants and funding

This work was supported by the Science and Technology Research Program of Chongqing Municipal Education Commission (Nos. KJZD-M202200801, KJQN202200828 and KJQN201800807). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.