
Deep Q Network (DQN) DQN은 Reinforcement Learning(RL)에 Deep Learning(DL)을 성공적으로 결합한 최초의 Deep Reinforcement Learning(DRL) 알고리즘으로 [Playing Atari with D

Double DQN : decoupling action select and evaluation

ReferencesPRIORITIZED EXPERIENCE REPLAY (Schaul et al., 2015)Pattern Recognition and Machine Learning (2006, Christopher Michael Bishop)Importance Sam

References PRIORITIZED EXPERIENCE REPLAY (Schaul et al., 2015) Importance Sampling in DRL Prioritized Experience Replay 기존 Deep Q Network(DQN)의 repl

Decoupling Q-value into state value and action advantage value
Outline기존 Value기반 방법은 각 state value function(혹은 state-action value function)을 찾아내서 그에따른 최적의 greedy policy를 찾아냈다.이러한 방법은 결정론적 정책을 사용함에 따른 한계가 명확했다. → a