Topology Adaptive Graph Convolutional Networks

Abstract

Spectral graph convolutional neural networks (CNNs) require approximation to the convolution to alleviate the computational complexity, resulting in performance loss. This paper proposes the topology adaptive graph convolutional network (TAGCN), a novel graph convolutional network defined in the vertex domain. We provide a systematic way to design a set of fixed-size learnable filters to perform convolutions on graphs. The topologies of these filters are adaptive to the topology of the graph when they scan the graph to perform convolution. The TAGCN not only inherits the properties of convolutions in CNN for grid-structured data, but it is also consistent with convolution as defined in graph signal processing. Since no approximation to the convolution is needed, TAGCN exhibits better performance than existing spectral CNNs on a number of data sets and is also computationally simpler than other recent methods.

13 pages

References in corpus (4)

• Very Deep Convolutional Networks for Large-Scale Image Recognition
• Discrete Signal Processing on Graphs
• Discriminative Embeddings of Latent Variable Models for Structured Data
• Spectral Projector-Based Graph Fourier Transforms

Cited by in corpus (40)

• Convolutional Neural Network Architectures for Signals Supported on Graphs
• Graph Neural Networks: Taxonomy, Advances and Trends
• End-to-End Differentiable Molecular Mechanics Force Field Construction
• Graph Signal Processing and Deep Learning: Convolution, Pooling, and Topology
• Hyper Meta-Path Contrastive Learning for Multi-Behavior Recommendation
• Machine-learned molecular mechanics force field for the simulation of protein-ligand systems and beyond
• Predicting Basin Stability of Power Grids using Graph Neural Networks
• Chemical-Reaction-Aware Molecule Representation Learning
• PolyGNN: Polyhedron-based Graph Neural Network for 3D Building Reconstruction from Point Clouds
• On the design space between molecular mechanics and machine learning force fields
• Toward Dynamic Stability Assessment of Power Grid Topologies using Graph Neural Networks
• Understanding the Message Passing in Graph Neural Networks via Power Iteration Clustering
• Distilling Knowledge from Graph Convolutional Networks
• Towards better Interpretable and Generalizable AD detection using Collective Artificial Intelligence
• CGNN: Traffic Classification with Graph Neural Network
• Graph Inference Learning for Semi-supervised Classification
• Guiding Cascading Failure Search with Interpretable Graph Convolutional Network
• TELESTO: A Graph Neural Network Model for Anomaly Classification in Cloud Services
• Forecaster: A Graph Transformer for Forecasting Spatial and Time-Dependent Data
• A Comparative Study of Rule Extraction for Recurrent Neural Networks
• Robust Node Classification on Graphs: Jointly from Bayesian Label Transition and Topology-based Label Propagation
• Semi-Supervised Node Classification by Graph Convolutional Networks and Extracted Side Information
• Room Classification on Floor Plan Graphs using Graph Neural Networks
• Optimization and Generalization Analysis of Transduction through Gradient Boosting and Application to Multi-scale Graph Neural Networks
• Perona: Robust Infrastructure Fingerprinting for Resource-Efficient Big Data Analytics
• Learning by Sampling and Compressing: Efficient Graph Representation Learning with Extremely Limited Annotations
• TactileSGNet: A Spiking Graph Neural Network for Event-based Tactile Object Recognition
• Optimization-Based Algebraic Multigrid Coarsening Using Reinforcement Learning
• A Unified Deep Learning Formalism For Processing Graph Signals
• Efficient Mixed Precision Quantization in Graph Neural Networks
• STD-Net: Structure-preserving and Topology-adaptive Deformation Network for 3D Reconstruction from a Single Image
• Comparisons of Graph Neural Networks on Cancer Classification Leveraging a Joint of Phenotypic and Genetic Features
• GSA-Forecaster: Forecasting Graph-Based Time-Dependent Data with Graph Sequence Attention
• GFCN: A New Graph Convolutional Network Based on Parallel Flows
• Dual GNNs: Graph Neural Network Learning with Limited Supervision
• Pooling in Graph Convolutional Neural Networks
• Graphs for deep learning representations
• Heuristic Semi-Supervised Learning for Graph Generation Inspired by Electoral College
• Curvature Graph Neural Network
• Bridging the Gap of AutoGraph between Academia and Industry: Analysing AutoGraph Challenge at KDD Cup 2020