Implementation of ARL(Adversarial Reinforcement Learning)-AIRL(Adversarial Inverse Reinforcement Learning) with the DDPG backbone. Basically, two discriminators are used to discriminate reward estimation and actual reward for one task and to discriminate the superior IRL policy with emerging RL policy for the other task.
highcansavci/reinforcement-learning-arl-airl
Implementation of ARL(Adversarial Reinforcement Learning)-AIRL(Adversarial Inverse Reinforcement Learning) with the DDPG backbone. Basically, two discriminators are used to discriminate reward estimation and actual reward for one task and to discriminate the superior IRL policy with emerging RL policy.
Python