High-dimensional coverage conjecture for full conformal quantile regression
Let ℓα(u)=αu\ell_{\alpha}(u)=\alpha uℓα(u)=αu for u≥0u\geq 0u≥0 and (α−1)u(\alpha-1)u(α−1)u for u<0u<0u<0 be the pinball loss, let η^n+1\hat{\eta}_{n+1}η^n+1 be the dual quantile-regression score, and let…