/teseDCSS

Trabalho do André Gonçalves

Primary LanguagePython

Playing DCSS with Deep Reinforcement Learning

The project consists in the bot that is able to play DCSS through Deep Reinforcement Learning.

The project was made to run under localy using the webtiles version of the game, the server used and the other settings can be defined in the config.toml file.

Running the project

To run the project the user must:

Dependencys

  • PyTorch

The MakeFile

To run the network, run the command make, to analyse the results, run the comand make res, after inserting all of the results in the folder results

Project Arquitecture

There are 4 main packages in the project:

Project Background

The GameConnection package is simplification of the Beem project, where we just use the code that gets us the game information suchs as JSON messages.

This project was made in colaboration with Universidade do Minho in the Informatics Department as a Masters Degrees Thesis in the area of Artificial Inteligence.