As Artificial Intelligence and Machine Learning still require a lot of researches, reading scientific papers is very important if we want to understand the fundamentals of methodologies, and notably, to get updated with the latest approaches. However, reading research papers is really a daunting task for many beginners, not only in AI but also in other research domains. So in this post, I would like to share with you my personal experiences as an ex-PhD student.
1. First of all, choose a repository or a search engine, e.g. scholar.google.com may be the most common site to search for academic articles.
2. Choose the optimal key words, it’s very important to limit the number of search results. For example if we want to find papers related to « how to apply deep learning in facial recognition » then just type « deep learning facial recognition ». Also think about limiting the range of published years because deep learning is quite recent, so papers about facial recognition from 20 years ago seem not to be relevant (unless you just want to make some references). In addition, there is a trick to quickly find more papers on the same topic: just add « overview » or « state of the art » or « summary » in your search keywords, you will probably find some papers that already make a summary about existing works with valuable references!
3. Then pick just a few papers of which the title and the abstract most correspond to your problem to read.
4. Print those papers to read. To me it’s very crucial. If we read papers on computers we will quickly get tired/bored and may be distracted by other things (social networks, email, … ). If you could, try to read papers outdoor or some relaxing space (I used to read papers in the coffee room of the lab when there was no one there). I mean: feel relax! You can listen to some kind of reading music etc if it helps you concentrate.
5. Use a highlighter pen to mark important points!
6. Read carefully the abstract and the conclusion first! This will ensure that you get the right papers to read. Sometimes I spent half a day to read a complicated paper then I realized that it was not the one I needed
7. For the main part: read slowly! Do not try to skip too many things because research papers are limited on the number of pages so they are very condensed: every word has its value.
8. After reading and understanding a paper, try to make a summary for the paper by yourself then store it for future references. Trust me, this step may be the most important one, because you will never want to read again and again a paper that you already read and understood in the past.
9. If there are something that you find unclear while reading, do not hesitate to email the authors as they are normally willing to answer questions related to their articles (like what I used to do).
10. Finally, do not forget to have a look at References section, because you may find some interesting papers of the same topic.
Hope you a very good research!
Thank you for reading & sharing.