Training a Gaming Agent on Brainwaves


This project involves the usage of the ErrP ERP as a reward to train artificial intelligence agents to solve basic games. Human subjects silently and quitely watch the agents play along the game, and by understanding the game itself, rewards signals are generated and used to train the agents using reinforcement learning.


The training of the agent was performed offline. The next step is to perform it online. Additionally, we want to verify other gaming strategies (the game is the simplest thing that we could found that triggered the ErrP signal)

  • Paper: