Learning Predictions for Algorithms with Predictions
This addresses the challenge of how to generate predictions for algorithm design, which is incremental as it builds on the existing paradigm of algorithms with predictions.
The paper tackles the problem of obtaining predictions for algorithms with predictions, especially in online settings, by introducing a general design approach that learns predictors and applies online learning techniques, demonstrating effectiveness on problems like bipartite matching and job scheduling with improvements over existing results and first learning-theoretic guarantees.
A burgeoning paradigm in algorithm design is the field of algorithms with predictions, in which algorithms can take advantage of a possibly-imperfect prediction of some aspect of the problem. While much work has focused on using predictions to improve competitive ratios, running times, or other performance measures, less effort has been devoted to the question of how to obtain the predictions themselves, especially in the critical online setting. We introduce a general design approach for algorithms that learn predictors: (1) identify a functional dependence of the performance measure on the prediction quality and (2) apply techniques from online learning to learn predictors, tune robustness-consistency trade-offs, and bound the sample complexity. We demonstrate the effectiveness of our approach by applying it to bipartite matching, ski-rental, page migration, and job scheduling. In several settings we improve upon multiple existing results while utilizing a much simpler analysis, while in the others we provide the first learning-theoretic guarantees.