Algorithms with Calibrated Machine Learning Predictions

TL;DR

This paper introduces calibration as a method to improve the integration of machine learning predictions into online algorithms, showing it enhances performance in ski rental and job scheduling problems through theoretical analysis and real-world data validation.

Contribution

It proposes using calibration for better uncertainty quantification in predictions, leading to improved algorithmic performance in practical online decision problems.

Findings

Calibrated advice outperforms other uncertainty methods in high-variance settings.

Using calibrated predictors significantly improves online job scheduling performance.

Theoretical and empirical validation confirms calibration's practical benefits.

Abstract

The field of algorithms with predictions incorporates machine learning advice in the design of online algorithms to improve real-world performance. A central consideration is the extent to which predictions can be trusted -- while existing approaches often require users to specify an aggregate trust level, modern machine learning models can provide estimates of prediction-level uncertainty. In this paper, we propose calibration as a principled and practical tool to bridge this gap, demonstrating the benefits of calibrated advice through two case studies: the ski rental and online job scheduling problems. For ski rental, we design an algorithm that achieves near-optimal prediction-dependent performance and prove that, in high-variance settings, calibrated advice offers more effective guidance than alternative methods for uncertainty quantification. For job scheduling, we demonstrate that using a calibrated predictor leads to significant performance improvements over existing methods. Evaluations on real-world data validate our theoretical findings, highlighting the practical impact of calibration for algorithms with predictions.