RCoCo: Contrastive Collective Link Prediction Across Multiplex Network in Riemannian Space

Link prediction typically studies the probability of future interconnectionamong nodes with the observation in a single social network. More often thannot, real scenario is presented as a multiplex network with common (anchor)users active in multiple social networks. In the literature, most existingworks study either the intra-link prediction in a single network or inter-linkprediction among networks (a.k.a. network alignment), and consider two learningtasks are independent from each other, which is still away from the fact. Onthe representation space, the vast majority of existing methods are built uponthe traditional Euclidean space, unaware of the inherent geometry of socialnetworks. The third issue is on the scarce anchor users. Annotating anchorusers is laborious and expensive, and thus it is impractical to work withquantities of anchor users. Herein, in light of the issues above, we propose tostudy a challenging yet practical problem of Geometry-aware Collective LinkPrediction across Multiplex Network. To address this problem, we present anovel contrastive model, RCoCo, which collaborates intra- and inter-networkbehaviors in Riemannian spaces. In RCoCo, we design a curvature-aware graphattention network (κ-GAT), conducting attention mechanism in Riemannianmanifold whose curvature is estimated by the Ricci curvatures over the network.Thereafter, we formulate intra- and inter-contrastive loss in the manifolds, inwhich we augment graphs by exploring the high-order structure of community andinformation transfer on anchor users. Finally, we conduct extensive experimentswith 14 strong baselines on 8 real-world datasets, and show the effectivenessof RCoCo.