Deep Recurrent Q-Learning for Partially Observable MDPs

TMats / survey

Summary of Paper Survey

https://tmats.github.io/survey/

16 stars 2 forks source link

Deep Recurrent Q-Learning for Partially Observable MDPs #146

Open TMats opened 6 years ago

TMats commented 6 years ago

Matthew Hausknecht, Peter Stone
Submitted on 23 Jul 2015 (v1), last revised 11 Jan 2017 (this version, v4)

TMats commented 6 years ago

DRQN(Deep Recurrent Q-Network)

DQNにLSTMを追加

DQNはatariで4フレーム使うことでMDPになる

For many games, a single game screen is insufficient to determine the state of the system. DQN infers the full state of an Atari game by expanding the state representation to encompass the last four game screens. Many games that were previously POMDPs now become MDPs.