This website requires JavaScript.

PersonaSAGE: A Multi-Persona Graph Neural Network

Gautam ChoudharyIftikhar Ahamath BurhanuddinEunyee KohFan DuRyan A. Rossi
Dec 2022
摘要
Graph Neural Networks (GNNs) have become increasingly important in recentyears due to their state-of-the-art performance on many important downstreamapplications. Existing GNNs have mostly focused on learning a single noderepresentation, despite that a node often exhibits polysemous behavior indifferent contexts. In this work, we develop a persona-based graph neuralnetwork framework called PersonaSAGE that learns multiple persona-basedembeddings for each node in the graph. Such disentangled representations aremore interpretable and useful than a single embedding. Furthermore, PersonaSAGElearns the appropriate set of persona embeddings for each node in the graph,and every node can have a different number of assigned persona embeddings. Theframework is flexible enough and the general design helps in the wideapplicability of the learned embeddings to suit the domain. We utilize publiclyavailable benchmark datasets to evaluate our approach and against a variety ofbaselines. The experiments demonstrate the effectiveness of PersonaSAGE for avariety of important tasks including link prediction where we achieve anaverage gain of 15% while remaining competitive for node classification.Finally, we also demonstrate the utility of PersonaSAGE with a case study forpersonalized recommendation of different entity types in a data managementplatform.
展开全部
图表提取

暂无人提供速读十问回答

论文十问由沈向洋博士提出,鼓励大家带着这十个问题去阅读论文,用有用的信息构建认知模型。写出自己的十问回答,还有机会在当前页面展示哦。

Q1论文试图解决什么问题?
Q2这是否是一个新的问题?
Q3这篇文章要验证一个什么科学假设?
0
被引用
笔记
问答