Reinforcement Learning for Stochastic Networks, Toulouse

Session

Parallel session: Reinforcement learning in MDPs with large state spaces

Jun 18, 2024, 1:30 PM

A001 (ENSEEIHT)

Description

Organizers and chairs: R. Srikant and Yashaswini Murthy

56. Joint learning and scheduling in queueing systems

Lei Ying (University of Michigan, Ann Arbor)

6/18/24, 1:30 PM

This talk presents our recent results on joint learning and scheduling in queueing systems.

62. Neural Inventory Control in Networks via Hindsight Differentiable Policy Optimization

Matias Alvo (Columbia Business School (Decision, Risk and Operations division))

6/18/24, 2:00 PM

Inventory management offers unique opportunities for reliably evaluating and applying deep reinforcement learning (DRL). We introduce Hindsight Differentiable Policy Optimization (HDPO), facilitating direct optimization of a policy's hindsight performance using stochastic gradient descent. HDPO leverages two key elements: (i) an ability to backtest any policy's performance on a sample of ...

66. A Doubly Robust Approach to Sparse Reinforcement Learning

Assaf Zeevi (columbia university)

6/18/24, 2:30 PM

We propose a new regret minimization algorithm for episodic sparse linear Markov decision process (SMDP) where the state-transition distribution is a linear function of observed features.
The only previously known algorithm for SMDP requires the knowledge of the sparsity parameter and oracle access to a reference policy.
We overcome these limitations by combining the doubly robust method...