Reinforcement Learning(Q-Learing) implemented Wechat Jump, Q-Learning实现的微信跳一跳程序(iOS版)
Requirements
- conda
- facebook-wda
- Image
- ImageDraw
- python 3
- iOS code sign certificate
Performance
The program is still training data on my computer. After 50 trials training, it can achive max score 20. So what's next?
After 1200 trials training, it can achive max score 170.
May 10 thousands or more trials training would make a better effect. But this's a basic version you could derivative to start.
Optimizting list
- Learning method could be optimized.
- Center location of target box could be optimized.
- Strategy of getting maxiumn score on one box could be optimized.
- etc.