Determinable and interpretable network representation for link prediction DOI Creative Commons
Yue Deng

Scientific Reports, Journal Year: 2022, Volume and Issue: 12(1)

Published: Oct. 20, 2022

As an intuitive description of complex physical, social, or brain systems, networks have fascinated scientists for decades. Recently, to abstract a network's topological and dynamical attributes, network representation has been prevalent technique, which can map substructures (like nodes) into low-dimensional vector space. Since its mainstream methods are mostly based on machine learning, black box input-output data fitting mechanism, the learned vector's dimension is indeterminable elements not interpreted. Although massive efforts cope with this issue included, say, automated learning by computer theory mathematicians, root causes still remain unresolved. Consequently, enterprises need spend enormous computing resources work out set model hyperparameters that bring good performance, business personnel finds difficulties in explaining practical meaning. Given that, from physical perspective, article proposes two determinable interpretable node methods. To evaluate their effectiveness generalization, Adaptive Interpretable ProbS (AIProbS), network-based utilize representations link prediction. Experimental results showed AIProbS reach state-of-the-art precision beyond baseline models some small whose distribution training test sets usually unified enough perform well. Besides, it make trade-off precision, determinacy (or robustness), interpretability. In practice, contributes industrial companies without but who pursue during early stage development require high interpretability better understand carry business.

Language: Английский

Synchronization Mechanism for Controlled Complex Networks under Auxiliary Effect of Dynamic Edges DOI Open Access
Lizhi Liu, Zilin Gao, Yi Peng

et al.

Electronics, Journal Year: 2024, Volume and Issue: 13(10), P. 1990 - 1990

Published: May 20, 2024

The scope of complex dynamical networks (CDNs) with dynamic edges is very wide, as it composed a class realistic including web-winding systems, communication networks, neural etc. However, classic research topic in CDNs, the synchronization control problem, has not been effectively solved for CDNs edges. This paper will investigate emergence mechanism from perspective large-scale systems. Firstly, CDN conceptualized an interconnected coupled system edge subsystem (ES) and node (NS). Then, based on proposed improved directed matrix ES model expanded inequality, this overcomes limitations coupling term design models strong correlation tracking targets between nodes Due to effect synthesized controller auxiliary ES, state can be realized CDN. Finally, through simulation examples, validity advantages our work compared existing methods are demonstrated.

Language: Английский

Citations

0
A Gradient-based Topology Design Algorithm Applied to Second-order Network for Pursuing Best Possible Synchronization* DOI

Huanjing Geng,

Xiaoli Li

2022 37th Youth Academic Annual Conference of Chinese Association of Automation (YAC), Journal Year: 2024, Volume and Issue: unknown, P. 618 - 623

Published: June 7, 2024

Language: Английский

Citations

0
Chaotic dynamics and synchronization under tripartite couplings: Analyses and experiments using single-transistor oscillators as metaphors of neural dynamics DOI
Ludovico Minati, Laura Sparacino, Luca Faes

et al.

Chaos Solitons & Fractals, Journal Year: 2024, Volume and Issue: 189, P. 115567 - 115567

Published: Oct. 18, 2024

Language: Английский

Citations

0
Simplicial motif predictor method for higher-order link prediction DOI
Rongmei Yang, Bo Liu, Linyuan Lü

et al.

Expert Systems with Applications, Journal Year: 2024, Volume and Issue: unknown, P. 126284 - 126284

Published: Dec. 1, 2024

Language: Английский

Citations

0
Determinable and interpretable network representation for link prediction DOI Creative Commons
Yue Deng

Scientific Reports, Journal Year: 2022, Volume and Issue: 12(1)

Published: Oct. 20, 2022

As an intuitive description of complex physical, social, or brain systems, networks have fascinated scientists for decades. Recently, to abstract a network's topological and dynamical attributes, network representation has been prevalent technique, which can map substructures (like nodes) into low-dimensional vector space. Since its mainstream methods are mostly based on machine learning, black box input-output data fitting mechanism, the learned vector's dimension is indeterminable elements not interpreted. Although massive efforts cope with this issue included, say, automated learning by computer theory mathematicians, root causes still remain unresolved. Consequently, enterprises need spend enormous computing resources work out set model hyperparameters that bring good performance, business personnel finds difficulties in explaining practical meaning. Given that, from physical perspective, article proposes two determinable interpretable node methods. To evaluate their effectiveness generalization, Adaptive Interpretable ProbS (AIProbS), network-based utilize representations link prediction. Experimental results showed AIProbS reach state-of-the-art precision beyond baseline models some small whose distribution training test sets usually unified enough perform well. Besides, it make trade-off precision, determinacy (or robustness), interpretability. In practice, contributes industrial companies without but who pursue during early stage development require high interpretability better understand carry business.

Language: Английский

Citations

2