Reinforcement Learning for Stochastic Networks, Toulouse

Session

Parallel session: Online learning in stochastic networks

Jun 17, 2024, 3:30 PM

A002 (ENSEEIHT)

Description

Organizers and chairs: Lei Ying and Weina Wang

33. Online Learning and Optimization for Queues with Unknown Demand Curve and Service Distribution

Xinyun Chen (The Chinese University of Hong Kong, Shenzhen)

6/17/24, 3:30 PM

We investigate an online learning and optimization problem in a queueing system having unknown arrival rates and service-time distribution. The service provider’s objective is to seek the optimal service fee $p$ and service capacity $\mu$ so as to maximize the cumulative expected profit (the service revenue minus the capacity cost and delay penalty). We develop an online learning algorithm is...

48. Recent Advances in Average-Reward Restless Bandits

Weina Wang (Carnegie Mellon University)

6/17/24, 4:00 PM

We consider the infinite-horizon, average reward restless bandit problem. For this problem, a central challenge is to find asymptotically optimal policies in a computationally efficient manner in the regime where the number of arms, N, grows large. Existing policies, including the renowned Whittle index policy, all rely on a uniform global attractor property (UGAP) assumption to achieve...

80. Scalable Learning in Weakly Coupled Markov Decision Processes

Chen Yan

6/17/24, 4:30 PM

We explore a general reinforcement learning framework within a Markov decision process (MDP) consisting of a large number $N$ of independent sub-MDPs, linked by global constraints. In the non-learning scenario, when the model meets a specific non-degenerate condition, efficient algorithms (i.e., polynomial in $N$) exist, achieving a performance gap smaller than $\sqrt{N}$ relative to the...