README.md

Dynamically Rule-Interposing Learning in Deep Reinforcement Learning

Here are the experiments of Dynamically Rule-Interposing Learning(DRIL) model.

For more information, please click each folder and read the README.

About DRIL

The main idea is about proposing a method that combines both knowledge representation and DRL.

References

[1] Haodi Zhang, Zihang Gao, Yi Zhou, et al. Faster and Safer Training by Embedding High-Level Knowledge into Deep Reinforcement Learning. arXiv preprint. 2019:1910.09986.