AI Glossary

Proximal policy optimization

Also called PPO

What does “Proximal policy optimization” mean?

Proximal policy optimization: A reinforcement-learning method that updates a policy while limiting changes that move too far from the prior policy.

Proximal policy optimization visual guide: A reinforcement-learning method that updates a policy while limiting changes that move too far from the prior policy.
Visual guide Open full size (new tab)

What is an example of “Proximal policy optimization”?

A trainer improves an agent's actions without allowing each update to change its behavior too abruptly.

Keep learning

See also

Reinforcement learningTermReward modelTermReward hackingTermDiscount factorTermExploration and exploitationTerm

Same topic

Q-learningTermTemporal difference learningTermPreference datasetTermQLoRATermPre-trainingTerm