HER

E441111

reinforcement learning technique

HER is a reinforcement learning technique that improves learning from sparse rewards by reinterpreting failed experiences as successful ones for alternative goals.

Try in SPARQL Jump to: Surface forms Statements Referenced by

Observed surface forms (1)

Surface form	Occurrences
Hindsight Experience Replay	0

Statements (42)

Predicate	Object
instanceOf	reinforcement learning technique ⓘ reinforcement learning technique ⓘ
abbreviation	HER NERFINISHED ⓘ
addresses	sparse reward problem ⓘ
aimsTo	improve learning stability in sparse reward settings ⓘ reduce sample complexity in goal-based tasks ⓘ
appliedTo	goal-conditioned reinforcement learning ⓘ multi-goal reinforcement learning ⓘ
assumes	environment with goal space ⓘ
benefits	navigation tasks with sparse rewards ⓘ robotic manipulation tasks ⓘ
category	off-policy data augmentation method ⓘ
commonlyCombinedWith	DDPG NERFINISHED ⓘ Deep Deterministic Policy Gradient NERFINISHED ⓘ
coreIdea	reinterpret failed experiences as successful ones for alternative goals ⓘ relabelling goals in past trajectories ⓘ
describedIn	paper "Hindsight Experience Replay" ⓘ
fullName	Hindsight Experience Replay NERFINISHED ⓘ
improves	learning from sparse rewards ⓘ sample efficiency in reinforcement learning ⓘ
inspired	extensions such as CHER and ARCHER ⓘ subsequent goal relabelling methods ⓘ
introducedBy	Alex Ray NERFINISHED ⓘ Bob McGrew NERFINISHED ⓘ Filip Wolski NERFINISHED ⓘ Jonas Schneider NERFINISHED ⓘ Josh Tobin NERFINISHED ⓘ Marcin Andrychowicz NERFINISHED ⓘ OpenAI researchers ⓘ Peter Welinder NERFINISHED ⓘ Rachel Fong NERFINISHED ⓘ
introducedIn	2017 ⓘ
modifies	stored transitions with alternative goals ⓘ
operatesOn	off-policy reinforcement learning algorithms ⓘ
publishedAt	NIPS 2017 NERFINISHED ⓘ Neural Information Processing Systems NERFINISHED ⓘ
relatedTo	goal relabelling ⓘ universal value function approximators ⓘ
requires	goal-conditioned reward function ⓘ
typeOf	model-free reinforcement learning enhancement ⓘ
usedIn	reinforcement learning ⓘ
uses	experience replay buffer ⓘ

Referenced by (1)

Full triples — surface form annotated when it differs from this entity's canonical label.

Hindsight Experience Replay → abbreviation → HER ⓘ