Learning-Augmented Online Scheduling with Parsimonious Preemption
Researchers have developed a new approach to online scheduling that integrates machine learning predictions to improve job latency while minimizing preemption. This work addresses the trade-off between algorithmic performance and the complexity of frequent preemptions, which are common in existing learning-augmented scheduling methods. The proposed algorithms achieve competitive latency bounds with a limited number of preemptions per job, even with imperfect predictions, extending the applicability of learning-augmented scheduling to more realistic scenarios. AI
IMPACT Introduces a novel algorithmic approach for online scheduling that leverages machine learning, potentially improving efficiency in resource allocation systems.