Inductive Relation Prediction by Disentangled Subgraph Structure

Currently,most existing inductive relation prediction approaches are based on subgraph structures,with subgraph features extracted using graph neural networks to predict relations.However,subgraphs may contain disconnected regions,which usually represent different semantic ranges.Because not all semantic information about the regions is helpful in relation prediction,we propose a relation prediction model based on a disentangled subgraph structure and implement a feature updating approach based on relevant semantic aggregation.To indirectly achieve the disentangled subgraph structure from a semantic perspective,the mapping of entity features into different semantic spaces and the aggregation of related semantics on each semantic space are updated.The disentangled model can focus on features having higher semantic relevance in the prediction,thus addressing a problem with existing approaches,which ignore the semantic differences in different subgraph structures.Furthermore,using a gated recurrent neural network,this model enhances the features of entities by sorting them by distance and extracting the path information in the subgraphs.Experimentally,it is shown that when there are numerous disconnected regions in the subgraph,our model outperforms existing mainstream models in terms of both Area Under the Curve-Precision-Recall(AUC-PR)and Hits@10.Experiments prove that semantic differences in the knowledge graph can be effectively distinguished and verify the effectiveness of this method.

Matching user identities across social networks with limited profile data

Privacy preservation is a primary concern in social networks which employ a variety of privacy preservations mechanisms to preserve and protect sensitive user information including age,location,education,interests,and others.The task of matching user identities across different social networks is considered a challenging task.In this work,we propose an algorithm to reveal user identities as a set of linked accounts from different social networks using limited user profile data,i.e,user-name and friendship.Thus,we propose a framework,ExpandUIL,that includes three standalone al-gorithms based on(i)the percolation graph matching in Ex-pand FullName algorithm,(i)a supervised machine learning algorithm that works with the graph embedding,and(ii)a combination of the two,ExpandUserLinkage algorithm.The proposed framework as a set of algorithms is significant as,(i)it is based on the network topology and requires only name feature of the nodes,(i)it requires a considerably low initial seed,as low as one initial seed suffices,(ii)it is iterative and scalable with applicability to online incoming stream graphs,and(iv)it has an experimental proof of stability over a real ground-truth dataset.Experiments on real datasets,Instagram and VK social networks,show upto 75%recall for linked ac-counts with 96%accuracy using only one given seed pair.