Snake - deep reinforcement learning

Used libraries

• tensorflow2
• gym_snake

Description

Deep Q Learning project aimed to play a game called "snake". The agent was self learning using 2 types of neural networks: classic neural networks (worked quite well) and convolutional neural networks (turned out to be a bad choice for the project).

After 3.5k games the average number of eaten apples was 10.37 (average of 50 games).

Game demo