Abstract:Message Passing Neural Networks (MPNNs) are a common type of Graph Neural Network (GNN), in which each node's representation is computed recursively by aggregating representations (messages) from its immediate neighbors akin to a star-shaped pattern. MPNNs are appealing for being efficient and scalable, how-ever their expressiveness is upper-bounded by the 1st-order Weisfeiler-Lehman isomorphism test (1-WL). In response, prior works propose highly expressive models at the cost of scalability and sometimes generalization performance. Our work stands between these two regimes: we introduce a general framework to uplift any MPNN to be more expressive, with limited scalability overhead and greatly improved practical performance. We achieve this by extending local aggregation in MPNNs from star patterns to general subgraph patterns (e.g.,k-egonets):in our framework, each node representation is computed as the encoding of a surrounding induced subgraph rather than encoding of immediate neighbors only (i.e. a star). We choose the subgraph encoder to be a GNN (mainly MPNNs, considering scalability) to design a general framework that serves as a wrapper to up-lift any GNN. We call our proposed method GNN-AK(GNN As Kernel), as the framework resembles a convolutional neural network by replacing the kernel with GNNs. Theoretically, we show that our framework is strictly more powerful than 1&2-WL, and is not less powerful than 3-WL. We also design subgraph sampling strategies which greatly reduce memory footprint and improve speed while maintaining performance. Our method sets new state-of-the-art performance by large margins for several well-known graph ML tasks; specifically, 0.08 MAE on ZINC,74.79% and 86.887% accuracy on CIFAR10 and PATTERN respectively.

Works for large-scale knowledge graph reasoning

Dynamically Pruned Message Passing Networks for Large-Scale Knowledge Graph Reasoning

River of No Return: Graph Percolation Embeddings for Efficient Knowledge Graph Reasoning

Attention-Based Relational Graph Convolutional Network for Knowledge Graph Reasoning

Knowledge Graph Reasoning Neural Network Based on Dynamic Programming

Graph Neural Prompting with Large Language Models

GK: A Graphormer Architecture Based Knowledge Graph Reasoning Model for Link Prediction

Relgraph: A Multi-Relational Graph Neural Network Framework for Knowledge Graph Reasoning Based on Relation Graph

Exploring & Exploiting High-Order Graph Structure for Sparse Knowledge Graph Completion

Learning Adaptive Propagation for Knowledge Graph Reasoning

Adaptive Graph Neural Network with Incremental Learning Mechanism for Knowledge Graph Reasoning

Neural Probabilistic Logic Learning for Knowledge Graph Reasoning

DAG Matters! GFlowNets Enhanced Explainer for Graph Neural Networks

Think-on-Graph: Deep and Responsible Reasoning of Large Language Model with Knowledge Graph

Learning Multi-Granularity and Adaptive Representation for Knowledge Graph Reasoning

KnowGNN: a knowledge-aware and structure-sensitive model-level explainer for graph neural networks

ReasoningLM: Enabling Structural Subgraph Reasoning in Pre-trained Language Models for Question Answering over Knowledge Graph

Deep Cognitive Reasoning Network for Multi-hop Question Answering over Knowledge Graphs.

Expanding the Scope: Inductive Knowledge Graph Reasoning with Multi-Starting Progressive Propagation

From Stars to Subgraphs: Uplifting Any GNN with Local Structure Awareness