ConceptAI & ML
Offline reinforcement learning
The task of learning a policy from a fixed historical dataset of interactions without interacting further with the environment.
Know first
Named as a prerequisite by the lectures that define this. The number is how many say so. A grey one has no page yet.
Reinforcement learning4Markov decision processes3Off-policy learning1Policies1Policy gradient1Policy training1Q-learning1Reinforcement learning basics1
Opens up
Concepts that name this one as a prerequisite.
Lecturers
5Each in their own terms, from the easiest level.
Emma Brunskill
Offline RL means learning a good decision policy from a fixed dataset without interacting further with the environment.Ashish Vaswani, Lucas Beyer & William Fedus
A form of reinforcement learning where the agent learns from a pre-collected dataset of interactions rather than through active trial-and-error.Sergey Levine
Developing reinforcement learning methods that can reuse previously collected data sets to create a data-driven RL framework without active online interaction.Aviral Kumar
Learning optimized policies to maximize reward using existing static interaction data without online trial and interaction with the environment.
Lectures that assume it
1Connections
15The lectures that define it or assume it, and the concepts next to it.
Concepts9
- Reinforcement learning
- Markov decision processes
- Off-policy learning
- Policy gradient
- Behavior Policy ($\pi_beta$)
- Cal-QL