This paper was accepted on the workshop on Regulatable ML at NeurIPS 2023.
Conformal Prediction (CP) is a technique of estimating threat or uncertainty when utilizing Machine Studying to assist abide by widespread Threat Administration rules typically seen in fields like healthcare and finance. CP for regression will be difficult, particularly when the output distribution is heteroscedastic, multimodal, or skewed. Among the points will be addressed by estimating a distribution over the output, however in actuality, such approaches will be delicate to estimation error and yield unstable intervals. Right here, we circumvent the challenges by changing regression to a classification downside after which use CP for classification to acquire CP units for regression. To protect the ordering of the continuous-output area, we design a brand new loss operate and current vital modifications to the CP classification methods. Empirical outcomes on many benchmarks exhibits that this easy method provides surprisingly good outcomes on many sensible issues.