graph neural networks

Graph neural networks (GNNs) remain challenging to understand and apply for secondary-school students and machine learning practitioners due to conceptual abstraction and fragmented pedagogical resources. Method: We propose a unified encoder–decoder pedagogical and practical framework that systematically integrates core GNN mechanisms—including message passing, GCN, and GAT—and designs task-specific decoders for node classification, link prediction, and other downstream tasks. Grounded in engineering practice, we develop the first reproducible GNN入门 (introductory) curriculum, combining theoretical exposition with large-scale homogeneous-graph experiments. Contribution/Results: We quantitatively characterize how model performance scales with training data size and graph structural complexity—revealing novel empirical patterns. The framework provides standardized benchmarking protocols, hyperparameter tuning guidelines, and task-adaptation strategies. It significantly enhances pedagogical interpretability and industrial deployability of GNNs, lowering barriers to entry without sacrificing technical rigor.

This study systematically investigates the effectiveness boundaries and practical value of Graph Neural Networks (GNNs) across twelve application domains. Building upon a unified design space, it derives both spectral and spatial formulations of GNNs from first principles, analyzes their expressive power through the lens of the Weisfeiler–Leman test, and evaluates domain-specific graph construction strategies and architectural choices. The work establishes the first cross-domain analytical framework that disentangles genuine performance gains from baseline biases, uncovering common challenges such as heterophily, scaling effects, and deployment gaps. It clarifies the applicability limits of GNNs, highlights the discrepancy between leaderboard-topping models and deployable ones, and offers constraint-aware practical guidelines to address issues including oversmoothing, over-squashing, and distributional shifts.

Traditional methods for node classification and clustering on graph-structured data—such as social and biological networks—are limited by their inability to effectively model non-Euclidean geometric properties inherent in graphs. To address this, we propose a synergistic modeling framework that integrates classical graph algorithms with graph neural networks (GNNs). Our approach systematically analyzes the representational disparities between these two paradigms and constructs an interpretable graph representation learning scheme, thereby providing theoretical foundations for non-Euclidean structural modeling. Extensive experiments on multiple benchmark graph datasets demonstrate that the proposed method achieves 43%–70% higher accuracy in both node classification and clustering compared to standalone classical algorithms (e.g., Label Propagation, Spectral Clustering) and baseline GNN models. These results robustly validate the dual advantages of our fusion strategy—superior accuracy and enhanced robustness—over existing approaches.

Existing studies lack quantitative theoretical analysis of the differences in optimization and generalization performance between graph neural networks (GNNs) and multilayer perceptrons (MLPs). Method: Leveraging feature learning theory, this paper provides the first rigorous analysis of two-layer graph convolutional networks (GCNs) trained via gradient descent, modeling ReLU^q activation functions and spectral properties of the expected degree matrix D, under a structure-guided signal-noise separation framework. Contribution/Results: We theoretically prove that graph convolution explicitly exploits graph structure to significantly enhance signal learning while suppressing noise memorization, thereby expanding the benign overfitting regime by approximately √D^{q−2} compared to CNNs. This constitutes the first quantitative characterization of the fundamental generalization advantage of GNNs over MLPs. Extensive simulations corroborate the superior generalization and robustness of GCNs predicted by our theory.

Existing graph Transformers suffer from limited effectiveness, poor scalability, and high preprocessing complexity, often failing to outperform simple GNNs. To address this, we propose the first pure-attention graph learning framework that treats edge sets—not nodes—as the fundamental modeling unit, eliminating conventional node-centric representations and hand-crafted message passing. Our method introduces vertically interleaved masked and standard self-attention encoders, coupled with attention-based pooling for end-to-end differentiable training. It requires no graph reconstruction or preprocessing, natively supports heterogeneous graphs and transfer learning. Evaluated across 70+ node- and graph-level benchmark tasks, our approach consistently surpasses tuned GNN baselines and state-of-the-art graph Transformers. It achieves new SOTA results on molecular graph classification, vision-based graph recognition, heterogeneous graph learning, and cross-domain transfer, while maintaining both high accuracy and linear scalability.

Existing deep learning models struggle to effectively encode spatial, topological, and semantic structural information inherent in images. This work systematically evaluates the impact of various visual graph construction strategies on image classification performance within a unified three-layer Graph Convolutional Network (GCN) framework. For the first time, it demonstrates that the graph structure itself plays a decisive role in model performance. The study underscores the critical importance of the graph construction preprocessing stage, providing empirical evidence that well-designed graph structures substantially enhance classification accuracy. These findings offer both methodological guidance and practical justification for graph structure selection and preprocessing in visual graph neural networks.

Existing graph representations—such as adjacency matrices—are incompatible with the text-processing paradigm of large language models (LLMs). To address this, we propose a reversible, locally structure-preserving mapping from graphs to instruction sequences: graphs are encoded into compact, deterministic instruction strings via adjacency matrix decomposition, yielding a parseable textual representation; a custom reversible parser enables lossless reconstruction. This work establishes the first bridge between graph algebra and LLM-native textual processing, simultaneously ensuring structural fidelity and sequence conciseness. Experiments demonstrate that our representation significantly improves LLM performance on graph modeling tasks—including graph classification and link prediction—validating both its effectiveness and generalizability across diverse graph domains.

This work investigates whether graph neural networks (GNNs) can exactly learn and execute classical graph algorithms under bounded-degree graph constraints and finite-precision arithmetic. To this end, the authors propose a two-stage approach: first training an ensemble of multilayer perceptrons (MLPs) to emulate the local computation rules of individual nodes, then embedding these MLPs into a GNN as its update functions to enable error-free inference. The study establishes the first theoretical framework for the learnability of graph algorithms within the LOCAL model of distributed computing and, leveraging neural tangent kernel (NTK) analysis, proves that algorithms such as message flooding, BFS, DFS, and Bellman-Ford can be learned and executed by GNNs with high probability and exact correctness from a small number of samples, thereby demonstrating the feasibility of GNNs to rigorously implement distributed graph algorithms.

Existing graph neural networks are constrained by their message-passing mechanisms, which struggle to efficiently capture long-range dependencies. This work proposes a linearized graph sequence modeling framework that, for the first time, reformulates information propagation on graphs as a sequence modeling problem. By decoupling computational depth from information propagation depth, the approach reveals the essential sequential properties required to preserve graph inductive biases. The method systematically integrates advanced sequence modeling paradigms with graph-structured priors, achieving substantial performance gains over current models across multiple long-range dependency tasks. These results demonstrate the effectiveness and superiority of leveraging sequence modeling to enhance graph representation learning.