Abstract:Graph convolutional network (GCN) nowadays become new state-of-the-art for networks representation learning. Most of the existing methods are single-granular methods that failed to analyze the graph at multi-granular views so as to lose abundant information. Advanced graph pooling techniques can be successfully benefiting from semi-supervised networks representation learning. How to capture multi-granular information through the graph pooling techniques on graphs without additional input features is a great challenge. Technically speaking, we propose our graph node embeddings framework, MGPOOL. First, inspired by the triadic influence learning, we use the 3-clique algorithm to coarsen the graph repeatedly. Three nodes of a triangle form a supernode. We treat the supernodes as key nodes for our graph pooling operations. That keeps the local relationship. These graphs capture consecutive 3-cliques from the finest to the coarsest to preserve global structural relationships. Second, we use the unsupervised single-granular algorithms on the coarsest graph to acquire its node embeddings. Based on that, our graph pooling operations combining with that node embeddings to generate another same size of the coarsest graph. This makes up for the uniqueness of the coarsening result at a time and expands the receptive field for each node to avoid high-proximity information lost. Third, we take the embeddings, the coarsest graph and new coarsest graph as uniform input of MGPOOL. We restore the coarsest graph to the original graph to get the original graph node embeddings. The experimental results on four public datasets, Wiki, Cora, CiteSeer, and DBLP, demonstrate that our method has a better Macro F1 value for node classification tasks and AUC and Ap value for link prediction than the baseline methods.

Maximal Independent Vertex Set applied to Graph Pooling

Maximal Independent Sets for Pooling in Graph Neural Networks

Uniform Pooling for Graph Networks

Pooling in Graph Convolutional Neural Networks

Learnable Pooling in Graph Convolution Networks for Brain Surface Analysis

SSPool: A Simple Siamese Framework for Graph Infomax Pooling

Graph Pooling Via Dual-view Multi-level Infomax.

Ipool - Information-based Pooling in Hierarchical Graph Neural Networks.

Mixed Pooling For Convolutional Neural Networks

Graph Pooling via Coarsened Graph Infomax

Unsupervised Hierarchical Graph Pooling Via Substructure-Sensitive Mutual Information Maximization

Multiactivation Pooling Method in Convolutional Neural Networks for Image Recognition.

Multi-scale Graph Pooling Approach with Adaptive Key Subgraph for Graph Representations

Graph Pooling for Graph Neural Networks: Progress, Challenges, and Opportunities

MxPool: Multiplex Pooling for Hierarchical Graph Representation Learning

Deep Generalized Max Pooling

Graph Cross Networks with Vertex Infomax Pooling

MaxCutPool: differentiable feature-aware Maxcut for pooling in graph neural networks

Channel Max Pooling For Image Classification

Graph explicit pooling for graph-level representation learning

MGPOOL: multi-granular graph pooling convolutional networks representation learning