arXiv:1905.07579
공식 코드 · Papers with Code
Combining Experience Replay with Exploration by Random Network Distillation
원문 초록
Our work is a simple extension of the paper "Exploration by Random Network Distillation". More in detail, we show how to efficiently combine Intrinsic Rewards with Experience Replay in order to achieve more efficient and robust exploration (with respect to PPO/RND) and consequently better results in terms of agent performances and sample efficiency. We are able to do it by using a new technique named Prioritized Oversampled Experience Replay (POER), that has been built upon the definition of what is the important experience useful to replay. Finally, we evaluate our technique on the famous Atari game Montezuma's Revenge and some other hard exploration Atari games.
함께 살펴볼 논문
공통 태그를 근거로 연결했습니다. 실제 인용 관계나 논문의 우수성을 뜻하지 않습니다.
Redeeming Intrinsic Rewards via Constrained Optimization
Reinforcement Learning with Latent Flow
Learning Abstract Models for Strategic Exploration and Fast Reward Transfer
Curiosity in Hindsight: Intrinsic Exploration in Stochastic Environments
On Bonus-Based Exploration Methods in the Arcade Learning Environment
Generalized Adaptive Transfer Network: Enhancing Transfer Learning in Reinforcement Learning Across Domains
토론
이 논문과 연결된 포럼 글이 아직 없습니다.