Learning to reinforcement learn
A repository for the purpose of recreating and expanding upon the experiments mentioned in Learning to Reinforcement Learn (Wang, et al. 2016) Specifically, this repository focuses on the implementation of a series of bandit problems (easy, medium, and hard) for generalization purposes. Additionally, Meta-RL enables the agent to continue learning, even while the weights are frozen.
Implementation inspired from Arthur Juliani, please see blog post for further details.