Playing Atari Games with Deep Reinforcement Learning and Human Checkpoint Replay
This addresses the challenge of exploration in reinforcement learning for sparse-reward environments, offering a domain-specific solution for Atari game playing.
The paper tackles the problem of learning to play difficult Atari games with sparse rewards by introducing human checkpoint replay, which uses checkpoints from human gameplay as starting points, resulting in substantial improvements over previous methods on games like Montezuma's Revenge and Private Eye.
This paper introduces a novel method for learning how to play the most difficult Atari 2600 games from the Arcade Learning Environment using deep reinforcement learning. The proposed method, human checkpoint replay, consists in using checkpoints sampled from human gameplay as starting points for the learning process. This is meant to compensate for the difficulties of current exploration strategies, such as epsilon-greedy, to find successful control policies in games with sparse rewards. Like other deep reinforcement learning architectures, our model uses a convolutional neural network that receives only raw pixel inputs to estimate the state value function. We tested our method on Montezuma's Revenge and Private Eye, two of the most challenging games from the Atari platform. The results we obtained show a substantial improvement compared to previous learning approaches, as well as over a random player. We also propose a method for training deep reinforcement learning agents using human gameplay experience, which we call human experience replay.