4SkyNet / hierarchical_IL_RL

Code for hierarchical imitation learning and reinforcement learning

Hierarchical Imitation - Reinforcement Learning

Code for our paper "Hierarchical Imitation and Reinforcement Learning"

Here you can find the implementation of Hierarchical DAgger, Hierarchical Behavior Cloning for the Maze Domain and Hybrid Imitation-Reinforcement Learning algorithms for the Atari game Montezuma's Revenge

Requires Tensorflow and Keras (the experiments were run on Tensorflow version 1.3.0 and Keras version 2.1.2. Note that I used Cuda version 8.0.61 and cuDNN 6.0.21)