How to read AI summit papers?

AI Most of my papers are in English , In particular, the top papers are in English . So how do we read it ？

I don't think English is a problem , Today's translation tools are very powerful , Basically, we can get the meaning of the thesis through translation .

The next question is how to read .

The chapters of the thesis are relatively fixed . There are mainly the following parts ：

• The first is the summary , A general introduction to the direction of the paper , Some of the advantages of the paper , If open source code , It will also be placed at the end of the summary .

• The first chapter is introduction , Introduce the research done by previous people and the author's thinking on these research deficiencies and the contribution of the paper , And achievements .

• The second chapter is related work , This part describes the contribution of the thesis in detail . Through reading this section, you can learn the advantages of the model .

• The third chapter describes the model in detail , This chapter contains not only formulas , It also describes the detailed structure of the model . This chapter is the core of the thesis .

• The fourth chapter is the experimental part , Introduce the author's parameter settings , Results achieved . This chapter has certain guiding significance for us to use model training .

• The fifth chapter is the summary , A summary of the achievements of the whole paper .

The number of chapters in a general paper is different , But they all have this structure , My reading order is as follows ：

• First step Look at the summary , After reading the abstract, you can see the direction of this paper , Know if this paper is what you want to find ？ The most important thing is to know if there is any source code .
• The second step Look at the profile , Find out how the author thinks about the problems he faces , The contribution of the paper .
• The third step See related work , After reading this , Have a more detailed understanding of the thesis .
• Step four Look at the experiment , If the paper is open source , Here you can try to do some reproduction work （ If conditions permit ）, You can run with your own data set , Feel the model .
• Step five Look at the summary .
• The last step Look at the models and formulas . Why should we put it at the end ？ This part is difficult to understand , The best way is to look at the source code . After we train on the model , Look slowly while waiting for the result . You can also debug the source code , Check whether the output result is consistent with the formula description .

The above is my way of reading the paper , I hope I can help you .