Transfer-Learning-In-RL

design a reinforcement learning algorithm that leverages prior experience to figure out how to solve new tasks quickly

How to run

Supported games:

• CartPole-v1 - cart
• MountainCarContinuous-v0 - car
• Acrobot-v1 - acro

Train single model

--mode=train --trained_model=cart --test_game=cart

Fine tune model for new game

--mode=fine --trained_model=acro --test_game=cart

Transfer learning using progressive network from 2 games to a new game

--mode=transfer --trained_model=cart --trained_model2=acro --test_game=car