Score-Aware Policy-Gradient Methods and Performance Guarantees using Local Lyapunov Conditions: Applications to Product-Form Stochastic Networks and Queueing Systems

no code implementations • 5 Dec 2023 • Céline Comte, Matthieu Jonckheere, Jaron Sanders, Albert Senen-Cerda

As a second contribution, we show that, under appropriate assumptions, the policy under a SAGE-based policy-gradient method has a large probability of converging to an optimal policy, provided that it starts sufficiently close to it, even with a nonconvex objective function and multiple maximizers.

Policy Gradient Methods Reinforcement Learning (RL)

Stochastic Non-Bipartite Matching Models and Order-Independent Loss Queues

no code implementations • 1 Dec 2020 • Céline Comte

The problem of appropriately matching items subject to compatibility constraints arises in a number of important applications.

Probability Performance 60K25 (Primary), 60J28 (Secondary) G.3