Career advise in AI/ML by Andrew Ng

Came across this video while random browsing and ended up watching all of it. Informative.
This is a talk delivered by Andrew Ng at Stanford on ‘How to read a research paper and Career advise in ML?’
Career advise part starts after 30 min. I found it very interesting and wish I had access to this information when I first graduated :slight_smile:
Note: Its a generic talk about making good decision and not necessarily about which topic is hot etc

Another good 2 page guidelines on How to Read a (research)paper:

Summary on Medium:

7 Likes

Thanks for sharing! Some running notes for section 1:

Reading 1 research paper in detail

Read it in multiple passes, instead of first word till last word.

Passes:

  1. Title + Abstract + Figures
  2. Pass (1) + Intro + Conclusion + Figures in more detail (+ Optional: Skim some sections)
  3. Pass (2) + Read all sections (just skim the math) (Ignore Related Works section)
  4. Read whole thing (can skips parts that are not understandable)

Depending upon the depth of understanding required for you, you can stop at any pass #.

Questions to ask yourself after reading:

  1. What did the authors try to accomplish?
  2. Key elements of the approach?
  3. How/Where it can be useful to you?
  4. Optional: References (papers) for in-depth or related follow-up.
  5. Optional: Discuss with peers

Sources of good papers

  1. Twitter
  2. ML-related sub-reddits
  3. Blogs (like medium)
  4. Top conferences
  5. Friends (Groups like Slack) / Online communities

Other tips:

  • To ensure if you really understood the math behind a math-intense paper, try re-deriving the math yourself after reading.
  • To ensure you understand the implementation of a paper:
    • Level 1: Just download and run the code
    • Level 2: Read through the approach
    • Level 3: Implement it in your own way
  • Do steady learning/reading, not short-bursts

Next section: ML Careeers (Video from 29:48)

5 Likes

@GokulNC
Will it be possible to create a separate category ‘Research’ (or a separate thread) for accumulating such information?
It will be nice to have a dedicated category for ‘Research’ related stuff if possible. Just a suggestion.

Good idea. Though I’m not sure if it’ll be helpful to create it as a separate category. (Since just ‘research’ is a very generic thing that might have not have good no. of topics if created as a separate section)

For example, if a person wants to discuss about a research paper in DL, they can do it under the DL section in General Discussions category. (and probably add a tag called research)

You can however create a thread for research in the AI Resources section; a thread that lists research related threads/articles. I’ll pin it to the top of the list or add to this: Learning Resources - Master List

Or please feel free to suggest how we can better structure the forum for research. :slight_smile:

You are right, I understand.

May be in future, if there are more posts, and I have some idea, I will share.
Thank you.

2 Likes