보정된 예측의 견고한 통합
이 논문은 개별적으로 calibrated된 여러 probabilistic forecasts를 통합하는 새로운 프레임워크를 제안합니다. 기존의 optimal-in-hindsight (OIH) benchmark가 달성 가능한 성능을 과소평가할 수 있음을 지적하며, calibration을 만족하는 모든 profile-wise conditional-mean mappings에 대해 보장할 수 있는 robust max-min benchmark를 도입합니다. 또한, forecast-only feedback 환경에서 robust benchmark를 달성하는 online algorithms를 제시합니다.
arXiv:2606.31020v1 Announce Type: new Abstract: Decision-makers often rely on multiple probabilistic forecasts that are individually calibrated but need not be fully informative. We develop a framework for aggregating such forecasts when the decision-maker knows only that experts satisfy calibration. We show that the joint distribution of calibrated forecasts can contain decision-relevant information that is unavailable from any single expert, so the standard optimal-in-hindsight (OIH) benchmark may substantially understate attainable performance. To formalize this idea, we introduce a robust max-min benchmark: the best payoff a decision-maker can guarantee against all profile-wise conditional-mean mappings compatible with calibration. This benchmark is tractable, admits a linear-programming formulation, and dominates the OIH benchmark up to calibration error. It can nevertheless be strictly below the Bayesian benchmark, clarifying the value of knowing experts' information structures. Finally, we provide online algorithms that attain the robust benchmark under forecast-only feedback and stronger contextual benchmarks under state feedback.