Skip to content

ConOLeary/Autometa

Repository files navigation

Final Year Project: Autometa

Report

  • See Diversifying_the METAs_of_Games_by_Analysing_the_Gameplay_Data_of_AI.pdf
  • MetaCaniX_Business_Plan.pdf is a write up of a business proposition that is centred around the technique developed in this project

training

High level concept

To train artificial intelligence to play games in an optimal manner, to analyse patterns in their play, and to tweak the game to diversify the choices made by these AI within the game.

Development Story

TLDR - the folder worth looking at is "Car Game". Pretty much all the development since March is in 'Car Game'. Before then I was testing what software would be optimal to manifest my high level concept through. 'Pong' was my first endeavour, where I discovered Javascipt is not an efficient language to do machine learning in. 'Shooter' was a HTML game I was making to manifest my concept through but which I dumped when I discovered pygame would be more appropriate for my ends. 'Frozen Lake' was something I was playing with when I was trying to wrap my head about deep q-learning. Finally, 'Car Game' was created when I had decided on what kind of game (car racing game) and what kind of software (eg Python, pygame) would be appropriate to manifest my concept through. Only during Car Game's development did I discover NEAT was the library that offered the appropriate machine learning approach to me, so there is still an outdated q-learning program for Car Game.

See Docs/development_log.txt for chronological development log.

About

Final Year Project: Auto-meta

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published