Conservation Laws, Extended Polymatroids and Multi-Armed Bandit Problems (Classic Reprint): An Unified Approach to Indexable Systems - Couverture souple
Synopsis
A unified framework for indexable systems in operations research
Explore how branching bandits, extended polymatroids, and multi-armed bandit ideas come together to solve complex scheduling and queueing problems. This book shows how to model projects as chains of jobs, apply generalized conservation laws, and derive practical indices to guide optimal decisions.
The text provides a clear path from theory to algorithms, revealing how discounted and undiscounted reward-tax problems can be analyzed with a consistent set of tools. It introduces the idea of generalized Gittins indices and explains when fixed priority rules are optimal, including concrete cases like Klimov’s problem and various multiclass queueing systems. Readers will see how to transform intricate control problems into tractable computations using extended polymatroids and index rules.
- Learn how to model project and job scheduling as branching bandits and see the resulting conservation laws in action.
- Understand generalized Gittins indices and their role in optimal policies for discounted and undiscounted problems.
- Discover algorithmic approaches (like Algorithm Ai) and how they produce optimal or near‑optimal policies.
- See applications to Klimov’s problem, multiclass queues, and deterministic scheduling without arrivals.
Ideal for researchers and practitioners who want a rigorous, single‑source approach to indexable systems and their wide range of applications.
Les informations fournies dans la section « Synopsis » peuvent faire référence à une autre édition de ce titre.
