Conformal Risk-Averse Decision Making with Optimized Certainty Equivalent Risk Control
Artificial IntelligenceInformation TheoryMachine Learning
Summary
The authors study how to make decisions when you're unsure about the real situation and want to avoid risky outcomes. They use a special measure called optimized certainty equivalent (OCE) to evaluate risk, which includes common measures like CVaR. When the situation's probabilities are known, they find an ideal way to act that links to prediction sets, helping explain some prediction methods. When probabilities are unknown, the authors suggest a new method using synthetic models and extra data to keep risks in check. They test their ideas on wireless beamforming problems.
risk-averse decision makingoptimized certainty equivalent (OCE)conditional value-at-risk (CVaR)conformal predictionprediction setsdata-driven calibrationwireless beamformingcontrol theoryuncertainty quantification
Authors
Amirmohammad Farzaneh, Osvaldo Simeone
Abstract
We study risk-averse decision making, in which an agent selects actions while being uncertain about the true system state. The risk is measured via optimized certainty equivalent (OCE) metrics, which generalize popular criteria such as mean-variance risk and conditional value-at-risk (CVaR). We characterize the optimal policy under known distributions, and show that it reduces to a prediction set-based solution for the CVaR. This provides an operational interpretation of conformal prediction-type prediction sets. For unknown distributions, we develop a data-driven calibration strategy, based on a synthetic model for the likelihood and held-out calibration data, yielding high-probability control of the OCE risk. The approach is evaluated on two wireless beamforming settings.