aac_kungfumaster

Kung-Fu Master with Advantage Actor-Critic

Implement a deep reinforcement learning agent for the Atari Kung-Fu Master Game and train it with Advantage Actor-Critic (AAC).

The agent is a convolutional neural network that converts states into action probabilities π and state values V.

Pre-processing

• Image resized to 42x42 and converted to grayscale to run faster
• Rewards divided by 100 'cuz they are all divisible by 100
• Agent sees last 4 frames of game to account for object velocity

Training on parallel games

To make actor-critic training more stable, you can play several games in parallel. To do this, initialize several parallel gym environments to which to send the agent's actions, and reset each environment if it terminates.