How to Read Research Papers !!

1. Compile list of papers(Medium/blog/GitHub/internet/ArXiX)
2. Skip around list (0%, 10%… 100%) Read or Understood.
3. 5-20 papers: Basic understanding
4. 50-100 papers: Very good understanding
5. 2 papers a week
6. Reading 1 paper (30 mins - 60mins) a. Read in Multiple passes i. Read title + abstract + figures. ii. Intro + conclusion + figures + skim rest + (skip/m related work). iii. Read but skip/skim maths. iv. Read whole thing but skip parts that don’t make sense. b. What did authors try to accomplish (Answer to yourself) c. What were the key elements of the approach? d. What can you use yourself? e. What other references do you want to follow?
7. Maths a. Re-derive the maths from scratch. (Learn how to derive own algorithms)
8. Code a. Run open-source code b. Re-implement from scratch
9. Steadily reading Sources of paper
10. Twitter
11. ML sub reddit
12. Top ML Conference: NIPS,ICML,KIR
13. Friends

Goal • Job [Big company or start-up] • PhD Most Important thing is: Do important work Recruiters look for: • Skills (ML, Quiz, Coding) • Meaningful work o Can do work • Broad area (10 + ) • Depth understanding of 1-2 area: Project, open source, research, internship.

Saturday Morning Problem • Read paper • Work on project/research • Open source • TV Selecting a Job • Work with great people/project

1. Focus on team (10-30 persons) you’ll interest with
2. Manager
3. Not on ‘brand’ • Learn the most. • Do important work.

AI Climate Landing AI EV AI