BriefGPT.xyz
Oct, 2017
Rainbow: 深度强化学习的综合改进
Rainbow: Combining Improvements in Deep Reinforcement Learning
HTML
PDF
Matteo Hessel, Joseph Modayil, Hado van Hasselt, Tom Schaul, Georg Ostrovski...
TL;DR
本文通过详尽的消融研究，考察了6种方法扩展DQN算法，实验结果表明这些方法的综合应用取得了Atari 2600基准测试最先进的性能，在数据效率和最终性能方面都取得了显著的改善。
Abstract
The
deep reinforcement learning
community has made several independent improvements to the
dqn algorithm
. However, it is unclear which of these extensions are complementary and can be fruitfully combined. This pa
→