Hybrid Reward Architecture for Reinforcement Learning · HackerPapers discussions

Hybrid Reward Architecture for Reinforcement Learning — mentioned by espadrine

Tue, 20 Jun 2017 20:33:49 GMT

The paper, “Hybrid Reward Architecture for Reinforcement Learning”[0], describes an evolution on DeepMind's Deep Q-Network (DQN) design[1] that was able to play many Atari games, presented in early 2015. The HRA design requires more preprocessing (DQN worked on the raw pixels and on score alone), but the charts mapping how much faster HRA learns and how much stronger it plays is convincing. [0]: https://arxiv.org/pdf/1706.04208.pdf [1]: https://deepmind.com/research/dqn/

Paper and discussion history. A mention is not an endorsement.

Hybrid Reward Architecture for Reinforcement Learning — HN submission

Mon, 19 Jun 2017 12:07:34 GMT

Hybrid Reward Architecture for Reinforcement Learning

3 points; 0 discussion replies on HN. Paper and source history. A submission is not an endorsement.

Hybrid Reward Architecture for Reinforcement Learning — mentioned by octoploid

Thu, 15 Jun 2017 10:09:21 GMT

Paper: https://arxiv.org/pdf/1706.04208.pdf

Paper and discussion history. A mention is not an endorsement.