Built a custom spaceship game to act as an environment for a D3QN Reinforcement learning Agent. Main goal being to tweak and mess around to gain understanding of Reinforcement Learning process as well increase knowledge of NN architecture for the noble purpose of just having more fun in the future. Written in python and mainly run in a conda env on windows vscode.
run main.py if you want to use args. List if arguments can be seen in file. (Will add here later)
run StarShip.py to play manually
random_policy.py runs the program with randomised inputs to simulate an Agent
DQNAgent.py can also be run on its own with its own default values and is equivalent to running main.py without any arguments
(Will clean up code later)
Papers/Articles i've found where useful(more coming):
https://nihit.github.io/resources/spaceinvaders.pdf https://towardsdatascience.com/activation-functions-neural-networks-1cbd9f8d91d6