how to read papers

i came across this video online by andrew ng where he talks about career advice and reading research papers and thought it was pretty insightful for those who are just starting to read academic papers. here is the link to the video.

according to andrew ng, he reads the paper in this following fashion and i hope to follow these steps and see if these work better than what i already do for myself (i used to just read top to bottom and this would take forever and ever).

⭐ do multiple passes through the paper ⭐ §

1. title/abstract/figures/tables
2. intro $\to$ conclusion $\to$ figures $\to$ skim the rest (the first time you read, skip through related work section)
3. read but skip math
4. read the whole thing but skip parts that don’t make sense

questions to ask yourself as you go through the steps above §

1. what did the authors try to accomplish?
2. what were the key elements of the approach?
3. what can you use it yourself?
4. what other references do you want to follow?

some of the resources for finding ML papers §

1. twitter
2. ML subreddit
3. NIPS/ICML/ICLR
4. arxiv sanity

UPDATE (12/2023): i recently discovered this thing called scholar-inbox and here is a tweet that talks about scholar inbox in more detail.

if you want to understand math that appears in papers: read $\to$ take detailed notes $\to$ re-derive it from scratch

if you want to be able to understand code that appears in papers: run open-source code or reimplement from scratch

learn steadily! don’t cram!