Multi-task Deep Reinforcement Learning with PopArt

12 Sep 2018Matteo HesselHubert SoyerLasse EspeholtWojciech CzarneckiSimon SchmittHado van Hasselt

The reinforcement learning community has made great strides in designing algorithms capable of exceeding human performance on specific tasks. These algorithms are mostly trained one task at the time, each new task requiring to train a brand new agent instance... (read more)

PDF Abstract

Results from the Paper


TASK DATASET MODEL METRIC NAME METRIC VALUE GLOBAL RANK RESULT LEADERBOARD
Atari Games Atari-57 PopArt-IMPALA Medium Human-Normalized Score 110.7% # 13
Visual Navigation Dmlab-30 PopArt-IMPALA Medium Human-Normalized Score 72.8% # 1