MIED : An Improved Graph Neural Network for Node Embedding in Heterogeneous Graphs

This paper proposes a Metapath-Infused Exponential Decay graph neural network (MIED) approach for node embedding in heterogeneous graphs. It is designed to address limitations existing methods, which usually lose the information during feature alignment and ignore different importance of nodes metapath aggregation. Firstly, convolutional (GCN) applied on subgraphs, derived from original with given metapaths transform features. Secondly, an exponential decay encoder (EDE) designed, influence starting point decays exponentially fixed parameter as they move farther away it. Thirdly, set experiments conducted two selected datasets graphs, i.e., IMDb DBLP, comparison purposes. The results show that MIED outperforms approaches, e.g., GAT, HAN, MAGNN, etc. Thus, our proven be able take full advantage considering weights based distance aspects. Finally, relevant parameters are analyzed recommended hyperparameter setting given.