discriminability of a graph filter

[[concept]]
Discriminability

The discriminability of a graph filter describes the ability of the filter to tell the difference between different eigenvalues of the shift operator spectrum.

Recall that we can think of the spectrum as a polynomial in the eigenvalues of the shift operator.

The discriminability thus corresponds with ∣∣h(λ)h(λ+ϵ)∣∣ , or the change in spectral response for a small change in eigenvalue.

Note

If a filter is able to discriminate well between eigenvalues, then it is able to describe which eigenvectors contribute to the task more effectively.

Mentions

File
GNNs perform better than their constituent filters
Lipschitz filters are stable to additive perturbations
convergence bound for graph convolutions
integral Lipschitz filters are stable to relative perturbations
integral lipschitz filters are stable to dilations
lipschitz graph filter
stability-discriminability tradeoff for Lipschitz filters
2025-03-05 graphs lecture 12
2025-03-10 graphs lecture 13
2025-04-07 lecture 18
2025-04-09 lecture 19