Total: 1
We propose a new method for finding the highest predictive density set or region using signed conformal inference. The proposed method is computationally efficient, while also carrying conformal coverage guarantees. We prove that under, mild regularity conditions, the conformal prediction set is asymptotically close to its oracle counterpart. The efficacy of the method is illustrated through simulations and real applications.