IMPALA

GPTKB entity

Statements (20)
Predicate Object
gptkbp:instanceOf gptkb:software
gptkbp:citation 1000+
gptkbp:developedBy gptkb:DeepMind
gptkbp:fullName INtegrated Model of Parallelism And Learning Algorithms
https://www.w3.org/2000/01/rdf-schema#label IMPALA
gptkbp:notableFeature V-trace off-policy correction
gptkbp:notablePublication Espeholt et al., 2018
gptkbp:openSource yes
gptkbp:releaseYear 2018
gptkbp:supports actor-critic methods
off-policy learning
scalable distributed RL
gptkbp:usedFor gptkb:reinforcement_learning
distributed training
gptkbp:usedIn Atari benchmarks
DMLab-30
DeepMind Lab
gptkbp:writtenBy gptkb:Python
gptkbp:bfsParent gptkb:RLlib
gptkbp:bfsLayer 7