REVIEW 3 cited by
Understanding convolution on graphs via energies
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
read the original abstract
Graph Neural Networks (GNNs) typically operate by message-passing, where the state of a node is updated based on the information received from its neighbours. Most message-passing models act as graph convolutions, where features are mixed by a shared, linear transformation before being propagated over the edges. On node-classification tasks, graph convolutions have been shown to suffer from two limitations: poor performance on heterophilic graphs, and over-smoothing. It is common belief that both phenomena occur because such models behave as low-pass filters, meaning that the Dirichlet energy of the features decreases along the layers incurring a smoothing effect that ultimately makes features no longer distinguishable. In this work, we rigorously prove that simple graph-convolutional models can actually enhance high frequencies and even lead to an asymptotic behaviour we refer to as over-sharpening, opposite to over-smoothing. We do so by showing that linear graph convolutions with symmetric weights minimize a multi-particle energy that generalizes the Dirichlet energy; in this setting, the weight matrices induce edge-wise attraction (repulsion) through their positive (negative) eigenvalues, thereby controlling whether the features are being smoothed or sharpened. We also extend the analysis to non-linear GNNs, and demonstrate that some existing time-continuous GNNs are instead always dominated by the low frequencies. Finally, we validate our theoretical findings through ablations and real-world experiments.
Forward citations
Cited by 3 Pith papers
-
Benchmarking Sheaf Neural Networks for Inductive Tasks
On 14 inductive graph benchmarks, sheaf neural networks underperform strong GNN baselines, and their performance is driven more by the surrounding architecture than by the sheaf diffusion mechanism.
-
TANGO: Graph Neural Dynamics via Learned Energy and Tangential Flows
TANGO adds a learnable energy gradient and an orthogonal tangential flow to GNN layers, improving long-range and heterophilic graph benchmarks.
-
Predicting Large-scale Urban Network Dynamics with Energy-informed Graph Neural Diffusion
A scalable spatiotemporal Transformer, ScaleSTF, matches the accuracy of much larger models on city-scale forecasting tasks at a fraction of the compute and memory cost.
Discussion (0). Sign in to comment.