Platform allowing easy analysis of DeepMind's algorithm DQN learning to play Mario Bros. Application provides neural network design tool and statistical tools for analysis.

Mario gameplay is emulated using FCEUX.

# First, install python dependencies and fceux emulator package:
sudo apt-get install python-pip python-dev python-tk build-essential git fceux
# Checkout this repository:
git pull https://github.com/kelsohmm/player-simulator.git
cd player-simulator
# Install required packages:
pip install -r requirements.txt
# Run application:
python gui_main.py

First, select an empty directory for you new session in the popup window. Next, choose your neural network architecture and hiperparameters.
Choose wisely - you can't change them later!

The session control window is divided into 3 sections

a) Left section
Allows for controlling algorithm runtime parameters and starting/stoping the learning session. Running the simulation will also open fceux emulator window

b) Center section
Displays overall statistics and charts for all played games.

c) Right section
Display statistics and charts for one selected game from history. Allows also for watching the game replay.