Hierarchical Graph Representation Learning with Differentiable Pooling

Abstract

Recently, graph neural networks (GNNs) have revolutionized the field of graph representation learning through effectively learned node embeddings, and achieved state-of-the-art results in tasks such as node classification and link prediction. However, current GNN methods are inherently flat and do not learn hierarchical representations of graphs---a limitation that is especially problematic for the task of graph classification, where the goal is to predict the label associated with an entire graph. Here we propose DiffPool, a differentiable graph pooling module that can generate hierarchical representations of graphs and can be combined with various graph neural network architectures in an end-to-end fashion. DiffPool learns a differentiable soft cluster assignment for nodes at each layer of a deep GNN, mapping nodes to a set of clusters, which then form the coarsened input for the next GNN layer. Our experimental results show that combining existing GNN methods with DiffPool yields an average improvement of 5-10% accuracy on graph classification benchmarks, compared to all existing pooling approaches, achieving a new state-of-the-art on four out of five benchmark data sets.

References in corpus (8)

• Inductive Representation Learning on Large Graphs
• Neural Message Passing for Quantum Chemistry
• Geometric deep learning: going beyond Euclidean data
• Learning Convolutional Neural Networks for Graphs
• Discriminative Embeddings of Latent Variable Models for Structured Data
• Predicting Organic Reaction Outcomes with Weisfeiler-Lehman Network
• Order Matters: Sequence to sequence for sets
• Graph Capsule Convolutional Neural Networks

Cited by in corpus (100)

• MeshCNN: A Network with an Edge
• Graph Self-Supervised Learning: A Survey
• Anomaly Detection on Attributed Networks via Contrastive Self-Supervised Learning
• Graph Neural Networks with convolutional ARMA filters
• A Comprehensive Survey on Deep Graph Representation Learning
• A Gentle Introduction to Deep Learning for Graphs
• Causal Attention for Interpretable and Generalizable Graph Classification
• Exploring Spatial Significance via Hybrid Pyramidal Graph Network for Vehicle Re-identification
• A Survey on Graph-Based Deep Learning for Computational Histopathology
• Differentiable Graph Module (DGM) for Graph Convolutional Networks
• Attention-based graph neural networks: a survey
• Understanding Pooling in Graph Neural Networks
• TodyNet: Temporal Dynamic Graph Neural Network for Multivariate Time Series Classification
• Second-Order Pooling for Graph Neural Networks
• Graph Kernels: A Survey
• Multilevel Graph Matching Networks for Deep Graph Similarity Learning
• Graph Rationalization with Environment-based Augmentations
• On Proximity and Structural Role-based Embeddings in Networks: Misconceptions, Techniques, and Applications
• GDPNet: Refining Latent Multi-View Graph for Relation Extraction
• Graph Neural Networks for Natural Language Processing: A Survey
• Graph Neural Networks for Text Classification: A Survey
• A Deep Generative Model for Graph Layout
• Graph Kernels: State-of-the-Art and Future Challenges
• Hierarchical Representation Learning in Graph Neural Networks with Node Decimation Pooling
• A Survey on Influence Maximization: From an ML-Based Combinatorial Optimization
• Attacking Graph Convolutional Networks via Rewiring
• Janossy Pooling: Learning Deep Permutation-Invariant Functions for Variable-Size Inputs
• DDGK: Learning Graph Representations for Deep Divergence Graph Kernels
• GraphVAMPNet, using graph neural networks and variational approach to markov processes for dynamical modeling of biomolecules
• AdaGNN: Graph Neural Networks with Adaptive Frequency Response Filter
• SE-GSL: A General and Effective Graph Structure Learning Framework through Structural Entropy Optimization
• Graph Pooling via Coarsened Graph Infomax
• Deep Learning of High-Order Interactions for Protein Interface Prediction
• Learning physical properties of anomalous random walks using graph neural networks
• Graph Convolutional Neural Networks with Node Transition Probability-based Message Passing and DropNode Regularization
• GALA: Graph Diffusion-based Alignment with Jigsaw for Source-free Domain Adaptation
• Semi-Supervised Hierarchical Graph Classification
• Relational Pooling for Graph Representations
• Graph Capsule Aggregation for Unaligned Multimodal Sequences
• Higher-order Clustering and Pooling for Graph Neural Networks
• Intent-aware Recommendation via Disentangled Graph Contrastive Learning
• KGNN: Harnessing Kernel-based Networks for Semi-supervised Graph Classification
• M-Evolve: Structural-Mapping-Based Data Augmentation for Graph Classification
• Clique pooling for graph classification
• A Novel Normalized-Cut Solver with Nearest Neighbor Hierarchical Initialization
• Graph Adversarial Training: Dynamically Regularizing Based on Graph Structure
• Implementing graph neural networks with TensorFlow-Keras
• Lifelong Learning of Graph Neural Networks for Open-World Node Classification
• HAHE: Hierarchical Attentive Heterogeneous Information Network Embedding
• Spectral Multigraph Networks for Discovering and Fusing Relationships in Molecules
• Fake News Detection in Social Media using Graph Neural Networks and NLP Techniques: A COVID-19 Use-case
• Doing More with Less: Overcoming Data Scarcity for POI Recommendation via Cross-Region Transfer
• Accurate, Low-latency, Efficient SAR Automatic Target Recognition on FPGA
• CLDG: Contrastive Learning on Dynamic Graphs
• Sequential Recommendation with Graph Neural Networks
• Graph Condensation for Graph Neural Networks
• Unsupervised Inductive Graph-Level Representation Learning via Graph-Graph Proximity
• Learning to Count Isomorphisms with Graph Neural Networks
• Search to Capture Long-range Dependency with Stacking GNNs for Graph Classification
• Projective Ranking-based GNN Evasion Attacks
• View-based Explanations for Graph Neural Networks
• COIN: Communication-Aware In-Memory Acceleration for Graph Convolutional Networks
• Single Image 3D Hand Reconstruction with Mesh Convolutions
• RPT: Toward Transferable Model on Heterogeneous Researcher Data via Pre-Training
• Deep Representation Learning for Social Network Analysis
• A Graph Isomorphism Network with Weighted Multiple Aggregators for Speech Emotion Recognition
• Learning Second-Order Attentive Context for Efficient Correspondence Pruning
• On Graph Classification Networks, Datasets and Baselines
• GNN-LoFI: a Novel Graph Neural Network through Localized Feature-based Histogram Intersection
• Bi-Level Graph Neural Networks for Drug-Drug Interaction Prediction
• Learning Tree Structures from Leaves For Particle Decay Reconstruction
• Zoo Guide to Network Embedding
• Graph Convolutional Neural Networks via Motif-based Attention
• Grouping-matrix based Graph Pooling with Adaptive Number of Clusters
• CARL-G: Clustering-Accelerated Representation Learning on Graphs
• Identifying Autism Spectrum Disorder Based on Individual-Aware Down-Sampling and Multi-Modal Learning
• Simplicial Complex Representation Learning
• The Impact of Global Structural Information in Graph Neural Networks Applications
• Deep Representations for Time-varying Brain Datasets
• Simplifying Clustering with Graph Neural Networks
• Heterogeneous Graph Matching Networks
• GraphAttacker: A General Multi-Task GraphAttack Framework
• DeSCo: Towards Generalizable and Scalable Deep Subgraph Counting
• Generalizing Downsampling from Regular Data to Graphs
• CCGG: A Deep Autoregressive Model for Class-Conditional Graph Generation
• Co-attention Graph Pooling for Efficient Pairwise Graph Interaction Learning
• Highly Accurate Disease Diagnosis and Highly Reproducible Biomarker Identification with PathFormer
• Maximum Entropy Weighted Independent Set Pooling for Graph Neural Networks
• Parameter Estimation for the SEIR Model Using Recurrent Nets
• Self-Parameterization Based Multi-Resolution Mesh Convolution Networks
• On Graph Neural Network Ensembles for Large-Scale Molecular Property Prediction
• Sparse hierarchical representation learning on molecular graphs
• Relational dynamic memory networks
• Interpretable Drug Synergy Prediction with Graph Neural Networks for Human-AI Collaboration in Healthcare
• Which Hyperparameters to Optimise? An Investigation of Evolutionary Hyperparameter Optimisation in Graph Neural Network For Molecular Property Prediction
• Layout-Graph Reasoning for Fashion Landmark Detection
• STR-GODEs: Spatial-Temporal-Ridership Graph ODEs for Metro Ridership Prediction
• A Structural Feature-Based Approach for Comprehensive Graph Classification
• Maximal Independent Sets for Pooling in Graph Neural Networks
• Weisfeiler and Leman go Hyperbolic: Learning Distance Preserving Node Representations