Minimax estimation of linear functionals in the Darcy inverse problem

Abstract

We study estimation of linear functionals $\int fψ$ of the conductivity $f$ in the Darcy problem $\nabla\cdot(f\nabla u)=g$ from observations of $u$ in Gaussian white noise of level $N^{-1/2}$. When the potential has a single nondegenerate minimum, we determine the minimax rates over Sobolev balls in the linearised experiment. Unless a compatibility condition along the gradient flow holds, the rate is governed by an effective dimension $T$ of the Hessian at the minimum, not by the ambient dimension: it is $N^{-(2s+T)/(4s+8)}$ for $T<4$, has an elbow at $T=4$ with an extra logarithmic factor, and is parametric for $T>4$. In the nonlinear experiment we show that these lower bounds persist when compatibility fails at the reference and $T\le4$, and we prove the minimax rate on classes with isotropic critical Hessians, and the parametric rate near references with $T>4$. In dimension at most four, no nonzero linear functional is uniformly $\sqrt N$-estimable on a Sobolev neighbourhood of the constant conductivity, even when it is regular there.