Nicolas Heess* Jonathan J Hunt* Timothy P Lillicrap David Silver
Google Deepmind
[Review & Implementation]
http://rll.berkeley.edu/deeprlworkshop/papers/rdpg.pdf
(1) Reinforcement Learning을 통해 풀고자 하는 문제들 중 "Partially Observable Environment"는 굉장히 어려운 문제입니다.
(2)
(1) model-free RL for continuous control deterministic policy gradient with recurrent neural network
(2) model-free RL for stochastic value gradient with recurrent neural network
LSTM (Long-Short-Term-Memory) 와 결합된 모델은 특히나 다양한 메모리를 요구하는 환경에서 뛰어난 성능을 보여주었습니다.
