Adversarial nonnegative matrix factorization for temporal link prediction

• Published in: Physics letters. A Vol. 527; p. 129984
• Main Authors: Zhang, Ting, Lv, Laishui, Bardou, Dalal
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 15-12-2024
• Subjects: Adversarial training; Network attacks; Nonnegative matrix factorization; Temporal link prediction; Network attacks; Adversarial training; Nonnegative matrix factorization; Temporal link prediction
• ISSN: 0375-9601
• DOI: 10.1016/j.physleta.2024.129984

Temporal link prediction has been extensively studied and widely applied in various applications, aiming to predict future network links based on the historical networks. However, most existing methods ignore the behavior of previous network updating information in temporal networks. To address these issues, we propose a novel link prediction model based on adversarial nonnegative matrix factorization, which fuses graph representation and adversarial learning to perform temporal link prediction. Specifically, we add a bounded adversary matrix to the input matrix to provide the robustness against real perturbations. Then, our model fully exploits the impact of snapshots by using communicability. Simultaneously, we utilize the cosine similarity to extract the node similarity and map it to low-dimensional latent representation to preserve the local structure. Additionally, we provide effective updating rules to learn the parameters of this model. Extensive experiments results on six real-world networks demonstrate that the proposed method outperforms several classical and the state-of-art matrix-based methods.
•Using communicability and cosine similarity to exploit the global and local structure of networks.•Adversarial NMF-based temporal link prediction approach is proposed.•Using Majorization-Minimization method to optimize the adversarial framework.•The new approach is applied to six real-world temporal networks and it outperforms other matrix-based methods.