PAC MAN
In this project, I took the pac-man version which I had built earlier, and used NEAT evolution algorithm to teach our pac-man to consume maximum possible dots while evading ghosts! Certain difficulties I faced were:
- determining the population size, and number of generations to run the NEAT algorithm for
- determining what inputs to provide the NEAT neural networks
Eventually, my pac-man trained on my local workstation, could consume half 3/4th of the total dots after being trained for 300 generations. I will continue to look into how to improve this statistic.
- python-neat : To run NEAT
- pygame : To develop the underlying pac-man game.