Measuring Over-smoothing beyond Dirichlet energy

• URL: http://arxiv.org/abs/2512.06782v1
• Date: Sun, 07 Dec 2025 10:53:22 GMT
• Title: Measuring Over-smoothing beyond Dirichlet energy
• Authors: Weiqi Guan, Zihao Shi,
• Abstract summary: We propose a family of node similarity measures based on the energy of higher-order feature derivatives.<n>We show that attention-based Graph Neural Networks (GNNs) suffer from over-smoothing when evaluated under these proposed metrics.
• Score: 0.0
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: While Dirichlet energy serves as a prevalent metric for quantifying over-smoothing, it is inherently restricted to capturing first-order feature derivatives. To address this limitation, we propose a generalized family of node similarity measures based on the energy of higher-order feature derivatives. Through a rigorous theoretical analysis of the relationships among these measures, we establish the decay rates of Dirichlet energy under both continuous heat diffusion and discrete aggregation operators. Furthermore, our analysis reveals an intrinsic connection between the over-smoothing decay rate and the spectral gap of the graph Laplacian. Finally, empirical results demonstrate that attention-based Graph Neural Networks (GNNs) suffer from over-smoothing when evaluated under these proposed metrics.

Related papers

• Symmetry-protected topology and deconfined solitons in a multi-link $\mathbb{Z}_2$ gauge theory [45.88028371034407]
  We study a $mathbbZ$ lattice gauge theory defined on a multi-graph with links that can be visualized as great circles of a spherical shell.<n>We show that this leads to state-dependent tunneling amplitudes underlying a phenomenon analogous to the Peierls instability.<n>By performining a detailed analysis based on matrix product states, we prove that charge deconfinement emerges as a consequence of charge-fractionalization.
  arXiv  Detail & Related papers  (2026-03-02T22:59:25Z)
• Analysis of Dirichlet Energies as Over-smoothing Measures [48.49843360392601]
  We analyze the distinctions between two functionals often used as over-smoothing measures.<n>We highlight critical distinctions necessary to select the metric that is spectrally compatible with the GNN architecture.
  arXiv  Detail & Related papers  (2025-12-10T18:17:33Z)
• Rethinking Oversmoothing in Graph Neural Networks: A Rank-Based Perspective [5.482832675034467]
  We show that rank-based metrics consistently capture oversmoothing, whereas energy-based metrics often fail.<n> Notably, we reveal that a significant drop in the rank aligns closely with performance degradation, even in scenarios where energy metrics remain unchanged.
  arXiv  Detail & Related papers  (2025-02-07T00:55:05Z)
• Quantum optical scattering by macroscopic lossy objects: A general approach [55.2480439325792]
  We develop a general approach to describe the scattering of quantum light by a lossy macroscopic object placed in vacuum.<n>We exploit the input-output relation to connect the output state of the field to the input one.<n>We analyze the impact of the classical transmission and absorption dyadics on the transitions from ingoing to outgoing s-polariton.
  arXiv  Detail & Related papers  (2024-11-27T17:44:29Z)
• Bridging Smoothness and Approximation: Theoretical Insights into Over-Smoothing in Graph Neural Networks [12.001676605529626]
  We explore the approximation theory of functions defined on graphs. We establish a framework to assess the lower bounds of approximation for target functions using Graph Convolutional Networks (GCNs) We show how the high-frequency energy of the output decays, an indicator of over-smoothing, in GCNs.
  arXiv  Detail & Related papers  (2024-07-01T13:35:53Z)
• Fermi's golden rule rate expression for transitions due to nonadiabatic derivative couplings in the adiabatic basis [0.0]
  We provide an analysis of the Fermi's golden rule (FGR) rate expression for nonadiabatic transitions between adiabatic states. The resulting rate expression includes quadratic contributions of NDC terms and their couplings to Franck-Condon modes.
  arXiv  Detail & Related papers  (2024-05-04T15:34:54Z)
• Understanding Oversmoothing in Diffusion-Based GNNs From the Perspective of Operator Semigroup Theory [12.327920883065238]
  This paper presents an analytical study of the oversmoothing issue in diffusion-based Graph Neural Networks (GNNs)<n>We rigorously prove that oversmoothing is intrinsically linked to the ergodicity of the diffusion operator.<n>Our experimental results reveal that this ergodicity-breaking term effectively mitigates oversmoothing measured by Dirichlet energy.
  arXiv  Detail & Related papers  (2024-02-23T13:44:57Z)
• Nonparametric Partial Disentanglement via Mechanism Sparsity: Sparse Actions, Interventions and Sparse Temporal Dependencies [58.179981892921056]
  This work introduces a novel principle for disentanglement we call mechanism sparsity regularization. We propose a representation learning method that induces disentanglement by simultaneously learning the latent factors. We show that the latent factors can be recovered by regularizing the learned causal graph to be sparse.
  arXiv  Detail & Related papers  (2024-01-10T02:38:21Z)
• Boundary theories of critical matchgate tensor networks [59.433172590351234]
  Key aspects of the AdS/CFT correspondence can be captured in terms of tensor network models on hyperbolic lattices. For tensors fulfilling the matchgate constraint, these have previously been shown to produce disordered boundary states. We show that these Hamiltonians exhibit multi-scale quasiperiodic symmetries captured by an analytical toy model.
  arXiv  Detail & Related papers  (2021-10-06T18:00:03Z)
• Out-of-time-order correlations and the fine structure of eigenstate thermalisation [58.720142291102135]
  Out-of-time-orderors (OTOCs) have become established as a tool to characterise quantum information dynamics and thermalisation. We show explicitly that the OTOC is indeed a precise tool to explore the fine details of the Eigenstate Thermalisation Hypothesis (ETH) We provide an estimation of the finite-size scaling of $omega_textrmGOE$ for the general class of observables composed of sums of local operators in the infinite-temperature regime.
  arXiv  Detail & Related papers  (2021-03-01T17:51:46Z)
• Localisation in quasiperiodic chains: a theory based on convergence of local propagators [68.8204255655161]
  We present a theory of localisation in quasiperiodic chains with nearest-neighbour hoppings, based on the convergence of local propagators. Analysing the convergence of these continued fractions, localisation or its absence can be determined, yielding in turn the critical points and mobility edges. Results are exemplified by analysing the theory for three quasiperiodic models covering a range of behaviour.
  arXiv  Detail & Related papers  (2021-02-18T16:19:52Z)
• A Dynamical Central Limit Theorem for Shallow Neural Networks [48.66103132697071]
  We prove that the fluctuations around the mean limit remain bounded in mean square throughout training. If the mean-field dynamics converges to a measure that interpolates the training data, we prove that the deviation eventually vanishes in the CLT scaling.
  arXiv  Detail & Related papers  (2020-08-21T18:00:50Z)

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.