What is Trust region policy optimisation (TRPO)?

Trust region policy optimisation (TRPO)

Trust region policy optimisation (TRPO): A policy gradient method that constrains how far each update may move the policy, the predecessor of PPO.

Related terms

Group relative policy optimisation (GRPO)
Proximal policy optimisation (PPO)
Combinatorial optimisation
Direct preference optimisation (DPO)
Generative engine optimisation (GEO)
Mesa-optimisation
Off-policy learning
On-policy learning

Browse the full glossary →