In this project, we have decided to explore the possibilities to apply a new approach in the reinforment learning which is based on the structure of most of the tabletop games. Based on those structure, several models or players would play a game/problem with others models or experts. By organizing a series of rounds, each players would received as input its current status, the status of the game and how well the other models/players have done. Based on that information, the model could perform ajustments on the following decisions over the problem in order to improve its behaviour.
We provided a first example of how to model this system and a problem to test it which can be used template for future developments.
-Daniel Lopez Lopez (Former B.Sc, University of A Coruna)
-Enrique Fernandez-Blanco (University of A Coruna)