Determine whether graph-Laplacian eigenvalue estimators can be optimally debiased

Determine whether a debiasing method for graph-Laplacian-based estimators of the eigenvalues of the limiting differential operator can achieve rates faster than $1/\sqrt{n v_\mu(h)}$ under smoothness assumptions, and prove whether such debiasing is impossible for general coarse PI measures so that $1/\sqrt{n v_\mu(h)}$ is minimax on that class.

Background

The paper’s variance bounds suggest that the rate 1/nvμ(h)1/\sqrt{n v_\mu(h)} may not be minimax for estimating eigenvalues of the limiting operator when additional smoothness is available. This motivates the possibility of debiasing graph-Laplacian-based estimators.

The authors conjecture that debiasing requires smoothness and therefore cannot work for arbitrary coarse PI measures. If correct, this would establish the coarse PI class as a natural class on which the displayed rate is minimax. Both the existence and the impossibility aspects remain unresolved.

References

The rates derived in this paper suggest that the rate $1/\sqrt{nv_\mu(h)}$ is not minimax for eigenvalue estimation, and that graph Laplacian-based estimators could be debiased. We conjecture that such a debiasing method requires some smoothness, and is therefore impossible for a general coarse PI measure, suggesting that the rate $1/\sqrt{nv_\mu(h)}$ becomes minimax on this larger class. This would constitute another advantage of considering the class of coarse PI measures, which would be in this sense the ``right'' class on which to study the estimator $\lambda_{\mu_n}h$. We leave this interesting question to future work.

Spectral stability of empirical metric-measure Laplacians  (2608.23150 - Divol, 24 Aug 2026) in Remark following Theorem 1, Section 1