About this document
Flow-Based Policy For Online Reinforcement Learning by e22194649 is a document available to read on EtoBox.
The document introduces FlowRL, a novel online reinforcement learning framework that combines flow-based policy representation with Wasserstein-2-regularized optimization to enhance policy expressiveness and performance. It addresses the challenges of aligning flow-based models with reinforcement learning objectives, enabling efficient policy learning while maintaining multimodality. Empirical evaluations demonstrate FlowRL
- Author
- e22194649
- Language
- EN