Generalization Bounds for Spectral GNNs via Fourier Domain Analysis
This work provides theoretical insights for practitioners designing spectral GNNs, though it appears incremental as it builds on existing Fourier analysis methods.
The authors tackled the problem of understanding generalization behavior in spectral graph neural networks by analyzing them in the graph Fourier domain, deriving data-dependent generalization bounds that are tighter in linear cases and correlate with generalization gaps on real graphs.
Spectral graph neural networks learn graph filters, but their behavior with increasing depth and polynomial order is not well understood. We analyze these models in the graph Fourier domain, where each layer becomes an element-wise frequency update, separating the fixed spectrum from trainable parameters and making depth and order explicit. In this setting, we show that Gaussian complexity is invariant under the Graph Fourier Transform, which allows us to derive data-dependent, depth, and order-aware generalization bounds together with stability estimates. In the linear case, our bounds are tighter, and on real graphs, the data-dependent term correlates with the generalization gap across polynomial bases, highlighting practical choices that avoid frequency amplification across layers.