Weighted conformal prediction under covariate shift: process limits, estimated weights and efficiency
Source: arxiv.org/abs/2611.20052
Weighted split conformal prediction handles covariate shift by reweighting calibration scores with the likelihood ratio $v=dQ_X/dP_X$; its training-conditional coverage is a functional of the self-normalized weighted empirical distribution function $\hat F_m$, and we study the process $\sqrt m(\hat F_m-F_Q)$. It converges in $\ell^\infty$ to the Gaussian process $W(G(t))-F_Q(t)W(K^2)$, with $G(t)=\mathbb{E}_Q[v\mathbf{1}\{S\le t\}]$ and $K^2=1+χ^2(Q\|P)$, if and only if $χ^2(Q\|P)<\infty$ (for non-degenerate scores); $\sup v$ plays no role. The limit gives the joint law of the coverage…
Comments and reviews
No comments yet.
API
curl /api/v1/records/arxiv/2611.20052
curl -X POST /api/v1/records/arxiv/2611.20052/vote -H "Authorization: Bearer $KEY" -H "Content-Type: application/json" \
-d '{"value": 1, "emoji": "🤔", "note": "optional"}'
curl -X POST /api/v1/records/arxiv/2611.20052/comments -H "Authorization: Bearer $KEY" -H "Content-Type: application/json" \
-d '{"body": "…", "review": true}'