Davis-Kahan Theorem

[[concept]]
Theorem (Davis-Kahan)

Let T and T be self-adjoint Hilbert-Schmidt operators with eigenspectra (λi,,φi) and (λi,φi) respectively ordered by eigenvalue magnitude. Then

||φiφi||π2||TT||d(λi,λi)

Where |||| is the operator norm and d(λi,λi) is defined as

d(λi,λi)min(minji(|λiλj|),minji(|λjλi|))

Call the first interior min d1 and the second d2

Example

20250402-graph.png
d(λi,λi) corresponds to the minimum "eigengap" for the closest eigenvalue for eigenvalue λi in its own spectrum.

Note

If d(,) is large, then this is OK for fast convergence. But if this is small, then we also need small ||TT|| to get fast convergence.

Let {Gn,Xn}(W,X). We can use the Davis-Kahan Theorem with TW (the limiting graphon shift operator) and TWn (induced graphon shift operator).

||φ(TWn)φi(TW)||π2||TWTWn||d(λi(TW),λi(TWn))

This is good because from lecture 15, we know that if Gn converges to W in the cut norm, then TWn converges to TW in L2 (cut norm Convergence in the cut norm implies convergence in L2). Thus we have convergence in the numerator!

Does this mean everything is OK?

Mentions

Mentions

Created 2025-04-09 Last Modified 2025-05-13