RL
Technology & Computing abbreviation · Artificial Intelligence
| Stands for | Reinforcement learning |
|---|---|
| Categories | Technology & Computing |
RL stands for Reinforcement learning, an artificial intelligence abbreviation.
A branch of machine learning in which an agent learns by trial and error, taking actions in an environment and receiving rewards or penalties, gradually discovering a strategy that maximizes long-term reward. It is used in game-playing AI, robotics and the fine-tuning of language models.
Example: The warehouse robot learned to grasp odd-shaped parcels through RL in simulation before touching real boxes.
Usage notes
Richard Sutton and Andrew Barto formalized much of the field and received the 2024 Turing Award for it. DeepMind's AlphaGo, which beat Lee Sedol at Go in 2016, combined RL with deep neural networks. Designing rewards is tricky, since agents exploit loopholes, a problem called reward hacking.
Related abbreviations
More artificial intelligence abbreviations
- PPO – Proximal policy optimization
- PR – Precision-recall (curve)
- QA – Question answering
- RAG – Retrieval-augmented generation
- ReLU – Rectified linear unit
- RF – Random forest
- RLAIF – Reinforcement learning from AI feedback
- RLHF – Reinforcement learning from human feedback
- RMSE – Root mean square error
- RNN – Recurrent neural network
- ROC – Receiver operating characteristic
- RPA – Robotic process automation