Flappy Bird Clone and Reinforcement Learning bot

FlapPyBird [1] is a Flappy Bird clone and the bot [2] learns via Q-Learning. I thought it's better to play it alongside the bot, pretty sure you don't need me to tell you which side is me. Well it's embarrassing. I tried, really. I never played the original, so I couldn't say how much is the same, but this clone does look good and sound nice. The bot did learn from mistakes, I think it reused data, because when I first ran it, it ran into obstacles quite a few times, but as of the recording, it performs perfectly. And please don't tell me the bot has faster game speed, you don't need to demoralize me further. I also don't know anything about *learning (RL, deep, neural, ...) you see popping out one a day. This probably is the first one I ever ran, but if you are interested the bot's README has howitworks and some graphs. $$$$ Window size is 288x512. $$$$ FlapPyBird * Cloned since 2014-04-03 * git-4fc78596c9f4abf0071a5490734cee1c33fd4fa0 (2014-08-03) * In Python with Pygame * By Sourabh Verma * MIT License flappybird-qlearning-bot * git-1e991b2f1bc1353c40487c235a1c6db81e86f227 (2016-06-01) * In Python with Pygame * MIT License [1]: https://github.com/sourabhv/FlapPyBird [2]: https://github.com/chncyhn/flappybird-qlearning-bot

Видео Flappy Bird Clone and Reinforcement Learning bot автора Начинаем С JS: Видеокурс