Transfer Reinforcement Learning

This is the supplementary files for final project of BU EC500 K1 / CS591 S2 Deep Learning, Transfer Reinforcement Learning. The A3C algorithms in this repo is based on @miyosuda.

1. Obtain the expert RL agent

run a3c.py in example folder. You can change games' name in the script.

2. Perform policy regression

run either actor_mimic.py or learn_from_demo.py in example folder. You can change task id in the script for different kinds of task.

3. Retrain A3C agent

run either retrain_a3c.py in example folder. You can change task id in the script for different kinds of task.