This article was reproduced to Know ID: Charles (Bai Lu) knows his personal column

Download the W3Cschool Mobile App, 0 Foundation Anytime, Anywhere Learning Programming >> Poke this to learn

Yesterday, I found this interesting thing when I looked at the higher open source projects for deep learning on GitHub:

Use Deep Intensified Learning to crack the Flipy Bird game (Deep Q-Learning).

Related documents

Baidu web download link: https://pan.baidu.com/s/19LgDHq0V3IpE1K5sfuug2g

Password: tqus

References

The content is mainly referenced from the GitHub open source project:

Using Deep Q-Network to Learn How To Play Flappy Bird

link:

https://github.com/yenchenlin/DeepLearningFlappyBird

Introduction to the principle

This project refers to the deep Q learning algorithm in deep enhancement learning and shows that this learning algorithm can be extended to crack The Flipy Bird game. That is, the project is trained using variations of Q-learning, whose input is that the original pixel output is a numerical function of the estimated action.

PS：

If you are interested in in-depth intensive learning, a paper called Demystifying Deep Reinforcement Learning is also available in the Public Number document, which is highly recommended by the original author.

Network architecture:

Prior to this, the pretreatment was:

(1) Grayscale image;

(2) Image size adjustment to 80×80;

(3) Stack every 4 frames into an 80x80x4 input array.

The network's final output is a matrix of 2×1 to determine whether the bird is moving. (i.e. whether to press the screen . . .

Test the environment

Computer system: Win10

Python version: 3.5.4

Python-related third-party libraries:

TensorFlow_GPU version: 1.4.0

Pygame version: 1.9.3

OpenCV-Python version: 3.3.0

For configuration details, please refer to the relevant network documentation!!!

Run the demo

The command-line window enters the DeepLearningFlappyBird folder to enter py -3.5 deep_q_network.py return to run:

The results are as follows:

More references

（1） Mnih Volodymyr, Koray Kavukcuoglu, David Silver, Andrei A. Rusu, Joel Veness, Marc G. Bellemare, Alex Graves, Martin Riedmiller, Andreas K. Fidjeland, Georg Ostrovski, Stig Petersen, C harles Beattie, Amir Sadik, Ioannis Antonoglou, Helen King, Dharshan Kumaran, Daan Wierstra, Shane Legg, and Demis Hassabis. H uman-level Control through Deep Reinforcement Learning. Nature, 529-33, 2015.

（2） Volodymyr Mnih, Koray Kavukcuoglu, David Silver, Alex Graves, Ioannis Antonoglou, Daan Wierstra, and Martin Riedmiller. P laying Atari with Deep Reinforcement Learning. NIPS, Deep Learning workshop.

（3）Kevin Chen. D eep Reinforcement Learning for Flappy Bird Report | Youtube result.

link:

https://youtu.be/9WKBzTUsPKc

（4）https://github.com/sourabhv/FlapPyBird

（5） https://github.com/asrivat1/DeepLearningVideoGames