Learning-Augmented Ski Rental with Discrete Distributions: A Bayesian Approach
By: Bosun Kang, Hyejun Park, Chenglin Fan
Potential Business Impact:
Helps computers decide best when they guess wrong.
We revisit the classic ski rental problem through the lens of Bayesian decision-making and machine-learned predictions. While traditional algorithms minimize worst-case cost without assumptions, and recent learning-augmented approaches leverage noisy forecasts with robustness guarantees, our work unifies these perspectives. We propose a discrete Bayesian framework that maintains exact posterior distributions over the time horizon, enabling principled uncertainty quantification and seamless incorporation of expert priors. Our algorithm achieves prior-dependent competitive guarantees and gracefully interpolates between worst-case and fully-informed settings. Our extensive experimental evaluation demonstrates superior empirical performance across diverse scenarios, achieving near-optimal results under accurate priors while maintaining robust worst-case guarantees. This framework naturally extends to incorporate multiple predictions, non-uniform priors, and contextual information, highlighting the practical advantages of Bayesian reasoning in online decision problems with imperfect predictions.
Similar Papers
Competitive Algorithms for Cooperative Multi-Agent Ski-Rental Problems
Machine Learning (CS)
Helps groups decide fairly when sharing costs.
Calibrating Geophysical Predictions under Constrained Probabilistic Distributions
Atmospheric and Oceanic Physics
Improves computer weather forecasts using known patterns.
Calibrating Geophysical Predictions under Constrained Probabilistic Distributions
Atmospheric and Oceanic Physics
Improves weather forecasts by learning from past patterns.