betaFeedback

WorksMathematics

Weighted conformal prediction under covariate shift: process limits, estimated weights and efficiency

Abstract

Weighted split conformal prediction handles covariate shift by reweighting calibration scores with the likelihood ratio v=dQX/dPXv=dQ_X/dP_X; its training-conditional coverage is a functional of the self-normalized weighted empirical distribution function F^m\hat F_m, and we study the process m(F^m−FQ)\sqrt m(\hat F_m-F_Q). It converges in ℓ∞\ell^\infty to the Gaussian process W(G(t))−FQ(t)W(K2)W(G(t))-F_Q(t)W(K^2), with G(t)=EQ[v1{S≤t}]G(t)=\mathbb{E}_Q[v\mathbf{1}\{S\le t\}] and K2=1+χ2(Q∥P)K^2=1+χ^2(Q\|P), if and only if χ2(Q∥P)<∞χ^2(Q\|P)<\infty (for non-degenerate scores); sup⁡v\sup v plays no role. The limit gives the joint law of the coverage across levels, consistent bootstraps, inflations asymptotically exact uniformly in the level, and Berry-Esseen-type rates. With estimated ratios, a uniform expansion remains valid when the estimate uses the calibration covariates; the estimation error enters through a covariance term that can reduce the variance below the oracle's, and a pairs bootstrap reproduces it. When the ratio is known, the weighted conformal threshold is semiparametrically efficient if and only if the ratio and the coverage indicator are uncorrelated under the test distribution, with an explicit efficient correction otherwise. When χ2(Q∥P)=∞χ^2(Q\|P)=\infty, the ratio has a regularly varying tail of index p∈(1,2)p\in(1,2) and the score decouples from it in the far tail, the finite-dimensional limits are pp-stable at the rate m1−1/pm^{1-1/p} up to a slowly varying factor.

Paper details

vlnc:W_9uamubm4
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
arXiv nonexclusive license
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

Claude Fable 5.1 (bare), GPT-6 Astra

View statement
Looking for related papers…