Abstract:Data in the form of graphs are prevalent, ranging from biological and social networks to citation graphs and the Web. In particular, most real-world graphs are heterogeneous, containing objects of multiple types, which present new opportunities for many problems on graphs. Consider a typical proximity search problem on graphs, which boils down to measuring the proximity between two given nodes. Most earlier studies on homogeneous or bipartite graphs only measure a generic form of proximity, without accounting for different “semantic classes”—for instance, on a social network two users can be close for different reasons, such as being classmates or family members, which represent two distinct semantic classes. Learning these semantic classes are made possible on heterogeneous graphs through the concept of metagraphs. In this study, we identify metagraphs as a novel and effective means to characterize the common structures for a desired class of proximity. Subsequently, we propose a family of metagraph-based proximity, and employ a learning-to-rank technique that automatically learns the right parameters to suit the desired semantic class. In terms of efficiency, we develop a symmetry-based matching algorithm to speed up the computation of metagraph instances. Empirically, extensive experiments reveal that our metagraph-based proximity substantially outperforms the best competitor by more than 10 percent, and our matching algorithm can reduce matching time by more than half. As a further generalization, we aim to derive a general node and edge representation for heterogeneous graphs, in order to support arbitrary machine learning tasks beyond proximity search. In particular, we propose the finer-grained anchored metagraph, which is capable of discriminating the roles of nodes within the same metagraph. Finally, further experiments on the general representation show that we can outperform the state of the art significantly and consistently across various machine learning tasks.

GraphMetaP: Efficient MetaPath Generation for Dynamic Heterogeneous Graph Models.

Long-range Meta-path Search on Large-scale Heterogeneous Graphs

GraphMSE: Efficient Meta-path Selection in Semantically Aligned Feature Space for Graph Neural Networks

Meta-multigraph Search: Rethinking Meta-structure on Heterogeneous Information Networks

Metapaths guided Neighbors aggregated Network for?Heterogeneous Graph Reasoning

Metagraph-Based Learning on Heterogeneous Graphs

Higher-Order Heterogeneous Graph Convolutional Network Based on Meta-Paths

Heterogeneous Graph Neural Networks Based on Meta-path

Differentiable Meta Multigraph Search with Partial Message Propagation on Heterogeneous Information Networks

Towards Scalable Subgraph Pattern Matching over Big Graphs on MapReduce.

Learning Heterogeneous Graph Embedding with Metapath-Based Aggregation for Link Prediction

Meta-path Reduction with Transition Probability Preserving in Heterogeneous Information Network

Simple and Efficient Metapath Aggregated Network for Recommendation

Meta-path-based heterogeneous graph neural networks in academic network

HMSG: Heterogeneous Graph Neural Network Based on Metapath Subgraph Learning.

Unsupervised Meta-path Reduction on Heterogeneous Information Networks

Meta-Path Constrained Random Walk Inference for Large-Scale Heterogeneous Information Networks

Reinforced Metapath Optimization in Heterogeneous Information Networks for Drug-Target Interaction Prediction

Megnn: Meta-path extracted graph neural network for heterogeneous graph representation learning

MetaNMP: Leveraging Cartesian-Like Product to Accelerate HGNNs with Near-Memory Processing