Reinforcement Learning RL

Learning by acting in an environment and adjusting behaviour based on rewards, rather than from labelled examples.

Machine Learning Intermediate 1 min read

Also known as: RL

Definition

In reinforcement learning an agent takes actions in an environment and receives rewards. There are no labelled examples; the agent has to work out which actions led to which outcomes, often when the reward arrives long after the action that caused it — the credit assignment problem.

The other defining tension is exploration versus exploitation: taking the action known to be decent, or trying something new that might be better.

RL is behind game-playing systems like AlphaGo, and — via RLHF — behind the tuning that makes chat models helpful rather than merely fluent.

How it relates to other terms

Narrower term RLHF Tuning a model using human preference comparisons, so it produces the kind of answer people actually rate high... Contrast with Supervised Learning Training on examples that come with the right answer attached, so the model learns to map inputs to known labe...
Referenced by
Supervised Learning Contrast with RLHF Narrower term

Related terms

Overfitting

When a model learns the training data so closely — including its noise — that it performs worse on anything new.

Beginner Machine Learning

Embedding

A list of numbers representing a piece of data, arranged so that similar things end up close together.

Intermediate Machine Learning

Neural Network

A model built from layers of simple weighted units, trained by adjusting those weights until the output matches the target.

Beginner Machine Learning

Was this definition helpful?

Definitions are reviewed by our editorial team. Spotted a problem? Tell us.

Keep reading

Browse the glossary Suggest a correction