In progress
- Neural net with input, 2xhidden and output layer
- Weights are optimized by a genetic algorithm
- Coming: Q-learning
- Obstacle detection: Vision in a radius of N pixels
- Food detection: 4 states, i.e. Ahead / Right / Behind / Left of Snake
Additionally, the Snake is given a hunger state, so it must eat within max_hunger steps or it dies.
Here: Vision-Radius = 2
Hunger death
- Design
- Q-Learning
- Documentation
- Show multiple generations
- pygame