Skip to main content

Reading Course

info

The purpose of reading course is three-fold. It's a safe and rewarding space for you to:

  1. Get up to speed on the research landscape/literature of your topic.
  2. Learn the highly-important, but often-neglected parts of research: stylistic scientific writing.
  3. Get deep into background knowledge necessary to do research on your topic.

Suggested format:

  1. NeurIPS LaTeX template, unlimited pages.
  2. Use the "Introduction" section to formulate and argue why your topic matters; not just in AI/ML but in real-world applications and impacts.
  3. Use the "Background" section to go as deep as possible in the background knowledge necessary for your topic. This will serve you well in the future!
    • E.g.: if the topic is "Bayesian Deep Learning", then:
      • start with probability theory---write the axioms, sigma-algebra, integration, Radon-Nikodym, etc. Basically, study a textbook, and summarize with your own words that textbook!
      • then Bayesian inference/stats and decision-making
      • then neural networks (architecture, optimization, probabilistic interpretation, etc.),
      • then the combination of the above.
  4. Use the "Literature Survey" to get up-to-speed with papers in your topics.
    • Start from the most well-known.
    • Then use the reference list to backtrace to older and older papers.
    • Use Google Scholar's citation tool to check newer papers that cite that paper.
  5. Optional, but you can also write a section or two to e.g.:
    • Do a meta-analysis (e.g., discussing patterns in the litearture),
    • Do an implementation/reproduction study. Very useful to get up to speed with the engineering side of the topic!
important

Again, it's important to note that this is a great chance for you to go deep into a topic; very early in your research career. What you learned in this course will compound greatly in the future!