Why does your graph neural network fail on some graphs? Insights from exact generalisation error

• URL: http://arxiv.org/abs/2509.10337v1
• Date: Fri, 12 Sep 2025 15:18:36 GMT
• Title: Why does your graph neural network fail on some graphs? Insights from exact generalisation error
• Authors: Nil Ayday, Mahalakshmi Sabanayagam, Debarghya Ghoshdastidar,
• Abstract summary: We derive the first exact generalisation error for a broad range of GNNs, including convolutional, PageRank-based, and attention-based models.<n>Our work provides a framework that explains when and why GNNs can effectively leverage structural and feature information.
• Score: 9.227034835080307
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Graph Neural Networks (GNNs) are widely used in learning on graph-structured data, yet a principled understanding of why they succeed or fail remains elusive. While prior works have examined architectural limitations such as over-smoothing and over-squashing, these do not explain what enables GNNs to extract meaningful representations or why performance varies drastically between similar architectures. These questions are related to the role of generalisation: the ability of a model to make accurate predictions on unlabelled data. Although several works have derived generalisation error bounds for GNNs, these are typically loose, restricted to a single architecture, and offer limited insight into what governs generalisation in practice. In this work, we take a different approach by deriving the exact generalisation error for GNNs in a transductive fixed-design setting through the lens of signal processing. From this viewpoint, GNNs can be interpreted as graph filter operators that act on node features via the graph structure. By focusing on linear GNNs while allowing non-linearity in the graph filters, we derive the first exact generalisation error for a broad range of GNNs, including convolutional, PageRank-based, and attention-based models. The exact characterisation of the generalisation error reveals that only the aligned information between node features and graph structure contributes to generalisation. Furthermore, we quantify the effect of homophily on generalisation. Our work provides a framework that explains when and why GNNs can effectively leverage structural and feature information, offering practical guidance for model selection.

Related papers

• Generalization of Graph Neural Networks is Robust to Model Mismatch [84.01980526069075]
  Graph neural networks (GNNs) have demonstrated their effectiveness in various tasks supported by their generalization capabilities. In this paper, we examine GNNs that operate on geometric graphs generated from manifold models. Our analysis reveals the robustness of the GNN generalization in the presence of such model mismatch.
  arXiv  Detail & Related papers  (2024-08-25T16:00:44Z)
• A Manifold Perspective on the Statistical Generalization of Graph Neural Networks [84.01980526069075]
  We take a manifold perspective to establish the statistical generalization theory of GNNs on graphs sampled from a manifold in the spectral domain.<n>We prove that the generalization bounds of GNNs decrease linearly with the size of the graphs in the logarithmic scale, and increase linearly with the spectral continuity constants of the filter functions.
  arXiv  Detail & Related papers  (2024-06-07T19:25:02Z)
• Towards Robust Fidelity for Evaluating Explainability of Graph Neural Networks [32.345435955298825]
  Graph Neural Networks (GNNs) are neural models that leverage the dependency structure in graphical data via message passing among the graph nodes. A main challenge in studying GNN explainability is to provide fidelity measures that evaluate the performance of these explanation functions. This paper studies this foundational challenge, spotlighting the inherent limitations of prevailing fidelity metrics.
  arXiv  Detail & Related papers  (2023-10-03T06:25:14Z)
• Debiasing Graph Neural Networks via Learning Disentangled Causal Substructure [46.86463923605841]
  We present a graph classification investigation on the training graphs with severe bias. We discover that GNNs always tend to explore the spurious correlations to make decision. We propose a general disentangled GNN framework to learn the causal substructure and bias substructure.
  arXiv  Detail & Related papers  (2022-09-28T13:55:52Z)
• MentorGNN: Deriving Curriculum for Pre-Training GNNs [61.97574489259085]
  We propose an end-to-end model named MentorGNN that aims to supervise the pre-training process of GNNs across graphs. We shed new light on the problem of domain adaption on relational data (i.e., graphs) by deriving a natural and interpretable upper bound on the generalization error of the pre-trained GNNs.
  arXiv  Detail & Related papers  (2022-08-21T15:12:08Z)
• Towards Self-Explainable Graph Neural Network [24.18369781999988]
  Graph Neural Networks (GNNs) generalize the deep neural networks to graph-structured data. GNNs lack explainability, which limits their adoption in scenarios that demand the transparency of models. We propose a new framework which can find $K$-nearest labeled nodes for each unlabeled node to give explainable node classification.
  arXiv  Detail & Related papers  (2021-08-26T22:45:11Z)
• Parameterized Explainer for Graph Neural Network [49.79917262156429]
  We propose PGExplainer, a parameterized explainer for Graph Neural Networks (GNNs) Compared to the existing work, PGExplainer has better generalization ability and can be utilized in an inductive setting easily. Experiments on both synthetic and real-life datasets show highly competitive performance with up to 24.7% relative improvement in AUC on explaining graph classification.
  arXiv  Detail & Related papers  (2020-11-09T17:15:03Z)
• Distance Encoding: Design Provably More Powerful Neural Networks for Graph Representation Learning [63.97983530843762]
  Graph Neural Networks (GNNs) have achieved great success in graph representation learning. GNNs generate identical representations for graph substructures that may in fact be very different. More powerful GNNs, proposed recently by mimicking higher-order tests, are inefficient as they cannot sparsity of underlying graph structure. We propose Distance Depiction (DE) as a new class of graph representation learning.
  arXiv  Detail & Related papers  (2020-08-31T23:15:40Z)
• Higher-Order Explanations of Graph Neural Networks via Relevant Walks [3.1510406584101776]
  Graph Neural Networks (GNNs) are a popular approach for predicting graph structured data. In this paper, we show that GNNs can in fact be naturally explained using higher-order expansions. We extract practically relevant insights on sentiment analysis of text data, structure-property relationships in quantum chemistry, and image classification.
  arXiv  Detail & Related papers  (2020-06-05T17:59:14Z)
• XGNN: Towards Model-Level Explanations of Graph Neural Networks [113.51160387804484]
  Graphs neural networks (GNNs) learn node features by aggregating and combining neighbor information. GNNs are mostly treated as black-boxes and lack human intelligible explanations. We propose a novel approach, known as XGNN, to interpret GNNs at the model-level.
  arXiv  Detail & Related papers  (2020-06-03T23:52:43Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.