MixHop: Higher-Order Graph Convolutional Architectures via Sparsified Neighborhood Mixing

Abstract

Existing popular methods for semi-supervised learning with Graph Neural Networks (such as the Graph Convolutional Network) provably cannot learn a general class of neighborhood mixing relationships. To address this weakness, we propose a new model, MixHop, that can learn these relationships, including difference operators, by repeatedly mixing feature representations of neighbors at various distances. Mixhop requires no additional memory or computational complexity, and outperforms on challenging baselines. In addition, we propose sparsity regularization that allows us to visualize how the network prioritizes neighborhood information across different graph datasets. Our analysis of the learned architectures reveals that neighborhood mixing varies per datasets.

References in corpus (1)

• Higher-order Graph Convolutional Networks

Cited by in corpus (43)

• Scaling Graph Neural Networks with Approximate PageRank
• GraphSAINT: Graph Sampling Based Inductive Learning Method
• Graph Representation Learning in Biomedicine
• Multipole Graph Neural Operator for Parametric Partial Differential Equations
• Graph Neural Networks: Taxonomy, Advances and Trends
• Graph Convolutional Networks for Graphs Containing Missing Features
• Connecting the Dots: Multivariate Time Series Forecasting with Graph Neural Networks
• K-Core based Temporal Graph Convolutional Network for Dynamic Graphs
• Regular Splitting Graph Network for 3D Human Pose Estimation
• Scalable Graph Neural Networks via Bidirectional Propagation
• Subgraph Neural Networks
• Bridging the Gap between Spatial and Spectral Domains: A Survey on Graph Neural Networks
• Path Integral Based Convolution and Pooling for Graph Neural Networks
• AdaGCN: Adaboosting Graph Convolutional Networks into Deep Models
• Interpreting and Unifying Graph Neural Networks with An Optimization Framework
• New Benchmarks for Learning on Non-Homophilous Graphs
• Sub-graph Contrast for Scalable Self-Supervised Graph Representation Learning
• Graph Random Neural Network for Semi-Supervised Learning on Graphs
• GCNH: A Simple Method For Representation Learning On Heterophilous Graphs
• A Network Science perspective of Graph Convolutional Networks: A survey
• Graph Representation Learning Beyond Node and Homophily
• Improving Graph Neural Networks with Simple Architecture Design
• Deep Graph Neural Networks with Shallow Subgraph Samplers
• Ripple Walk Training: A Subgraph-based training framework for Large and Deep Graph Neural Network
• Accurate, Efficient and Scalable Training of Graph Neural Networks
• SCE: Scalable Network Embedding from Sparsest Cut
• MONET: Debiasing Graph Embeddings via the Metadata-Orthogonal Training Unit
• SkipGNN: Predicting Molecular Interactions with Skip-Graph Networks
• Deoscillated Graph Collaborative Filtering
• Graph convolutions that can finally model local structure
• LookHops: light multi-order convolution and pooling for graph classification
• Dimensional Reweighting Graph Convolutional Networks
• Graph Highway Networks
• SoGCN: Second-Order Graph Convolutional Networks
• LEReg: Empower Graph Neural Networks with Local Energy Regularization
• Robust Hierarchical Graph Classification with Subgraph Attention
• Hybrid Low-order and Higher-order Graph Convolutional Networks
• Unsupervised Graph Representation by Periphery and Hierarchical Information Maximization
• Contrastive Adaptive Propagation Graph Neural Networks for Efficient Graph Learning
• Mutual Information Maximization in Graph Neural Networks
• Global Context Enhanced Social Recommendation with Hierarchical Graph Neural Networks
• Higher-Order Neighborhood Truss Decomposition
• DynACPD Embedding Algorithm for Prediction Tasks in Dynamic Networks