graph-neural-networks

This paper applies graph neural networks to predict the solvability of finite groups, demonstrating an AI-driven approach to a classic problem in group theory.

Proposes a topological framework for comparing trained Graph Neural Networks by mapping Stochastic Block Model embeddings onto the n-sphere, enabling visual inspection and transfer-learning candidate retrieval without retraining.

This paper proposes a spectral graph reinforcement learning framework for outage detection and power restoration in self-healing smart grids, achieving near-optimal real-time performance on IEEE test systems.

GiFlow is a graph-informed flow matching framework for spatiotemporal imputation that replaces Gaussian priors with a graph-informed prior, and uses a hybrid vector field model combining spatial attention, temporal attention, and spatiotemporal propagation. It outperforms state-of-the-art methods on synthetic and real-world datasets.

This paper proposes a framework that learns admissible cost partitions for planning heuristics by leveraging Lagrangian dual equivalence, using a deep architecture with axial self-attention to guarantee admissibility by construction. It claims to be the first machine-learned heuristic provably guaranteed to be admissible.

TCAR-Gen proposes a framework combining query-conditioned graph neural networks, temporal evidence fusion, and chain-of-trees reasoning for temporal graph retrieval in knowledge-grounded generation. It achieves improved recall on the Victorian Crime Diaries benchmark across multiple query types.

This paper analyzes how large language models internally process graph tokens in Graph Language Models (GLMs), finding a decoupling between activation-level saliency and graph-semantic utility. Graph sink tokens emerge as activation outliers but are not the primary carriers of graph structure, revealing limitations in current graph-token construction and alignment mechanisms.

Proposes GC-MoE, a graph-conditioned mixture of experts framework for traffic forecasting that assigns each node a personalized combination of frozen pretrained spatio-temporal GNN experts based on graph topology and recent input, training only a lightweight routing module (∼17K parameters) and achieving competitive performance on four benchmarks.

This paper presents the first model extraction attack on graph classification under strict black-box constraints, exploiting subgraph explanations to estimate decision boundaries. The findings reveal that mandated explainability interfaces create exploitable security vulnerabilities in Graph Neural Network services.

Introduces LoRe, a training-free wrapper that enforces per-step interaction budgets for iterative graph solvers, achieving substantial speedups and memory reductions on combinatorial optimization problems like MIS and TSP.

This paper proposes a modular temporal enhancement framework for signed graph neural networks that integrates historical context via a Historical Context Integration Module (HCIM) with LSTM and multi-head temporal attention, achieving consistent improvements on real-world temporal signed networks for dynamic link prediction.

This paper proposes CE-FedGNN, a federated graph neural network framework that achieves communication efficiency and privacy preservation by infrequently exchanging aggregated node representations with metric differential privacy guarantees, and demonstrates strong performance on benchmarks.

This book presents a comprehensive survey of graph theory under uncertainty, covering fuzzy, neutrosophic, and uncertain graph models, their properties, extensions, and applications in decision-making, graph neural networks, and knowledge graphs.

This paper presents a scalable heterogeneous graph neural network workflow for data-driven optimal power flow surrogate modeling, using distributed training on supercomputers and demonstrating improvements via fine-tuning pretrained models.

This paper explores the expressive power of Deep Homomorphism Networks (DHNs) for learning over relational databases, linking them to fragments of first-order logic and SQL, and analyzing static analysis problems like emptiness and subsumption.

This paper develops a systematic framework for establishing universality of machine learning models that handle inputs of varying dimensions (e.g., graphs with different node counts). It shows that many existing architectures fail to be universal and proposes simple modifications to restore universality.

WaveGraphNet is a coupled inverse-forward graph learning framework for guided-wave damage localization in composite plates, using sparse transducer networks and graph-based spectral descriptors to improve spatial generalization under limited training data.

This paper adapts instance discrimination self-supervised learning to link prediction in graphs, proposing new models L-GRACE and L-BGRL that operate on link representations and improve performance especially on unattributed graphs.

This paper introduces Transductive Sharpening (TS), a loss-level modification for semi-supervised node classification that minimizes prediction entropy on unlabeled nodes while counterbalancing on labeled nodes, achieving consistent performance improvements without architectural changes.

Introduces ArtifactLinker, a framework that models HuggingFace as an artifact graph and uses GNNs and LLM agents to automatically discover state-of-the-art models and research insights.