GrokFormer parameterizes graph spectral filters as Fourier series over higher-order Laplacian spectra, achieving state-of-the-art accuracy on many node and graph classification tasks.
As a result, GrokFormer learns to adaptively use a larger K to capture a broader range of frequency components, rather than restricting itself to a small K
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2024 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
GrokFormer: Graph Fourier Kolmogorov-Arnold Transformers
GrokFormer parameterizes graph spectral filters as Fourier series over higher-order Laplacian spectra, achieving state-of-the-art accuracy on many node and graph classification tasks.