A little experience sharing, many people shared the courses or papers they saw, but no one shared their own experience in systematic learning or scientific research. In fact, reproducing a paper is often able to see what the original author learned and made it For this thesis, a good systematic learning experience is very important. You can find a way to rise rapidly among many free knowledge, and the cost is your own time. The learning process is very important, just like the postgraduate entrance examination experience, but you need to seek truth from facts, and the one that suits you is the best.
theory
The following is a more systematic summary of my own intensive learning, and I also hope to provide a more systematic learning route based on personal experience. Free things require more tossing. Detours are essential, and not all courses are suitable for everyone. , everyone has their own basics, suitable for different levels of courses, 0-based intensive learning theory courses I personally recommend Mr. Wang Shusen's course, David Silver's course is in English, not very friendly, in fact, I read a lot when I entered the pit, Don't bother with python intensive learning, Xu Zhiqin of Beijing Jiaotong University, Southwest University of Science and Technology College of Optimal Control and Data Intelligence Team, Tang Yudi, also read some posts on Zhihu, such as vernacular intensive learning, some of the above are more practical , some of them are pure theory, and the videos can be found on bilibili. Personally, I think that the most easy-to-understand theory is the intensive learning course of Mr. Wang Shusen . It is recommended to watch and understand repeatedly. I also watched it several times before I understood. The principle of DDPG can help you understand related papers.
Practical
After mastering the theory, it needs to be consolidated with the code. Personally, I suggest mastering the tools first. Tensorflow has version 1.0 and version 2.0. Some of the codes downloaded from the Internet are version 1.0 and some are version 2.0. The tensorflow tutorial here personally recommends geek tensorflow1.0 [ Peking University] Tensorflow2.0 , these two are very friendly to novices, familiar with some variables, operations, tensorboard visualization, etc. (the last blog is the relevant course content). After having a certain foundation, next I recommend Zhihu’s Vernacular Intensive Learning , the relevant code address RLcode . Combining theory and tools can basically understand these codes and deepen them repeatedly. Next, I recommend Mo Fan python reinforcement learning, [Mo Fan Python] Intensive Learning Reinforcement Learning , although you may search for this at the beginning, but it is not recommended to watch this at the beginning, the video is short and concise, not suitable for those who have no foundation Classmates, after understanding Mr. Wang Shusen's theory, it will be very clear to watch Mofan's code explanation video.
summary
The course sequence link, I hope it can help you, fight monsters and upgrade in the field of reinforcement learning! ! !
- Course: Wang Shusen Intensive Learning Course Code: wangshusen/DeepLearning
- Course: Geek tensorflow1.0
- Course: [Peking University] Tensorflow2.0
- Course: Vernacular Reinforcement Learning Code: RLcode
- Course: [Don't bother with Python] Reinforcement Learning Code: MorvanZhou
/Reinforcement-learning-with-tensorflow