Pre-training with non-expert human demonstration for deep reinforcement learning

Explore this paper's citation graph

Summary

This work improves data efficiency in deep RL by addressing one of the two learning goals, feature learning, using supervised learning to pre-train on a small set of non-expert human demonstrations and empirically evaluate the approach using the asynchronous advantage actor-critic algorithms in the Atari domain.

Type
article
Published
2018-12-21
Cited by
27
References
45
Access
Open access

Keywords

Reinforcement learning, Leverage (statistics), Deep learning, Raw data, Feature learning

References

Cited by

Related papers

No related papers recorded.