A0966
Title: SStaGCN: Simplified stacking based graph convolutional networks
Authors: Jia Cai - Guangdong University of Finance and Economics (China) [presenting]
Abstract: Graph convolutional network (GCN) is a powerful model studied broadly in various graph structural data learning tasks. However, designing GCN models to mitigate the over-smoothing phenomenon is still a crucial issue to be investigated. A novel Simplified Stacking is proposed based on GCN (SStaGCN) by utilizing stacking ideas, aggrehich is a general adaptive framework for tackling distinct types of structural graph data. Specifically, we first use the base models of stacking to extract the node features in the graph. Subsequently, aggregation methods such as mean, attention and voting techniques are employed to enhance the ability of feature extraction further. Thereafter, the node features are considered as inputs and fed into the vanilla GCN model. Furthermore, theoretical generalization bound analysis of the proposed model is explicitly given. Extensive experiments on $3$ public citation networks and another $3$ heterogeneous tabular data demonstrate the effectiveness and efficiency of the proposed approach over several state-of-the-art GCNs. Notably, the proposed SStaGCN can efficiently mitigate the over-smoothing problem of GCNs.
