Integrity and Robust Network Embedding of Information Network with AAE

Bin LIU, Yun-fang CHEN, Wei ZHANG
2018 DEStech Transactions on Engineering and Technology Research  
Most existing network representation methods only consider network structure information or node content information independently, not making full use of them. Furthermore, noise that exists in real networks also causes the deterioration of the efficiency of traditional methods. In this paper, we propose IANE (Integrity Adversarial Network Embedding) framework, adversarial learning to regularize the representation learning in Adversarial AutoEncoders (AAE). IANE includes an information joint
more » ... information joint module and an adversarial learning module. The former aims to obtain complete input information, while the latter captures the highly non-linear structure of the network in the process of dimension reduction. Besides IANE has better robustness against noise by matching the posterior distribution of the embedding vectors to the given prior distribution. Our experiment results show that our method performs higher Macro-F1 than several baselines proposed in recent years.
doi:10.12783/dtetr/ecar2018/26358 fatcat:3y43nxcr6zbyzdgbhza2my3e64