10. ベースラインアルゴリズム
10
先読み検索の有無でベースラインを区別:
IRIS(提案手法)はMonte Carlo Tree Searchとの組み合わせが可能だが、
本論文では先読み検索なしの手法を比較対象として設定
先読み検索なし:
SimPLe [5]、CURL [6]、DrQ [7]、SPR [8]
先読み検索あり:
MuZero [9]、EfficientZero [10]
[5] Kaiser, Łukasz, et al. "Model Based Reinforcement Learning for Atari." 2019.
[6] Srinivas, Aravind, Michael Laskin, and Pieter Abbeel. "CURL: Contrastive Unsupervised Representations for Reinforcement Learning." 2020.
[7] Yarats, Denis, Ilya Kostrikov, and Rob Fergus. "Image augmentation is all you need: Regularizing deep reinforcement learning from pixels." 2020.
[8] Schwarzer, Max, et al. "Data-efficient reinforcement learning with self-predictive representations." 2020.
[9] Schrittwieser, Julian, et al. "Mastering atari, go, chess and shogi by planning with a learned model." 2020.
[10] Ye, Weirui, et al. "Mastering atari games with limited data." 2021.