Reinforcement learning
- In French:
- Apprentissage par renforcement
- Abbreviation:
- RL
Definition
A learning method in which an agent discovers by trial and error which actions bring it the most reward. It is not told the right action: it infers it from the consequences of its choices, which may appear long afterwards. In robotics, this training is often done in simulation, because trials on a real robot are slow, costly and sometimes dangerous.
Checked on against the references below.