Deep Reinforcement Learning with Double Q-learning

arxiv:1509.06461

Original source · Paper JSON

Reinforcement learning · 2015-09-22

Cites (0) · Cited by (12)

Complete cleaned Markdown · Source Markdown including bibliography

Read in parts

1. Part 1
2. Part 2
3. Part 3

Contents

• Deep Reinforcement Learning with Double Q-learning
• Abstract
• Background
• Deep Q Networks
• Double Q-learning
• Overoptimism due to estimation errors
• Theorem 1 .
• Double DQN
• Empirical results
• Results on overoptimism
• Quality of the learned policies
• Robustness to Human starts
• Discussion
• Acknowledgments
• Appendix
• Theorem 1 .
• Proof of Theorem 1.
• Theorem 2 .
• Proof.
• Experimental Details for the Atari 2600 Domain
• Network Architecture
• Hyper-parameters
• Supplementary Results in the Atari 2600 Domain
• Instructions for reporting errors