Announcement
A new reinforcement learning monograph examines data efficiency in tabular models
Its authors cover sample complexity across online, offline and generative-model learning, using model-based and model-free approaches.
TLDR
Submitted to arXiv on October 30, Statistical Foundations of Reinforcement Learning: A Non-Asymptotic Perspective examines how much data tabular reinforcement learning algorithms need. The six authors analyze learning with a generative model, online learning and offline learning. They also briefly discuss extensions including robust reinforcement learning and learning from human feedback.
Combined views
10.2K
3 Sources, first seen ago
