Recommended by friends , I saw B Stand in Teacher Liu er's class （ Video link ）, It's really good . This column is some of my notes after class .

although XXXXX, But not .

Overview

First of all, teacher Liu explained , Why do many university courses fall far behind the actual needs of enterprises , Why not use the latest ？
That's because these things update iteratively too fast , It's likely that you learned the framework routine when you were a freshman , By the time I was a sophomore or a junior, these things were no longer usable , So choose the framework in the application period , At this time, it is already in a mature state .（ In fact, it is the so-called master leading in , Practice in individual , The teacher takes you from 0-1, Then you are 1-0, still 1-∞, It depends on your ability ）,

• A language can be divided into ： Development period 、 Ice age 、 Application period .
  

Goals and requirements

• The goal is ： Skillfully use Pytorch; Understand neural networks and deep learning
• requirement ： Understand linear algebra and probability theory ;（ Recommend teacher Li Hang's statistical learning foundation 、 And the Bible Deep Learning）

Development

• One of the earliest AI Is a rule-based system , Previously used to calculate the original function .
• machine learning ： The so-called machine learning is to transform the process of human reasoning into algorithms , That is to replace the thinking of human brain with Algorithm ;
• Deep learning ： Taoists who have studied machine learning must know Feature Engineering , It's real trouble , More features , Affect over fitting , There are fewer features , Training is not fitted ,2333333… Then if we can solve the feature engineering step with Algorithm , Isn't that much more convenient ？ So for deep learning, we will train the simplest features , Then design a separate layer , Feature extraction , Then connect to the learner , And then output .

Here is a picture of the Bible , It introduces AI、 The concepts of machine learning and deep learning .

• Finally, I also talked about some partial derivatives , For the follow-up, we calculate the design of the graph . Won't it , It is suggested to turn over the book and study or switch packages .

Pytorch install

First enter the website link : pytorch. Pull down to find here

PyTorch Build Just choose the first one , Another one BETA Develop a test version
Package Is to choose what to install ,anconda perhaps pip Fine .
Compute Platform： It is suggested to choose CPU, Because you have to choose CUDA Words , The installation package is almost 1.3G, And the download speed is ridiculously slow .

Then there is the command line Input （ Cheating code ） Code. ,

There will be one to enter y The process of , Is to help you update other packages .
Then it's waiting …
Here is a Chinese version of the document , link : Chinese document .
Then you can listen to the class , Side facing documents and Github Programming the .