Background #

At the time of Jan 2023, I was a Masters student at the School of Computing(SoC) in NUS. I was in my second semester out of three total semesters, when I decided to take up a 6000 level module (mods).

SoC has 6 levels of modules where each module number indicates it’s average audience/students to be from. Here is a rough summary

However, there is no hard restrictions, i.e. you can enroll in a course above or below your recommended course. Since I had already taken three 5000 level mods in my first sem, I decided to take one 6000 level mod: CS6284: Advanced Topics in Reinforcement Learning taught by Prof. Lin Shao

Expectation vs Reality: 5000 vs 6000 level mod #

In every course I have taken previous the format was something like, the prof would come to class, give a lectures, and then you have to go back, study the materials yourself, do your assignments and finally take an exam or do a project by the end of the semester. However, for 6000 level mod, it’s different.

6000 level mods are seminar style mods. This means the major focus of the module is not just to teach the contents of the topics, but also train a student for their research careers. It involves readings research paper weekly, writing paper summaries, presenting some of the paper summaries in front of the class and maybe even do a course project.

CS6284: Advanced Topics in RL #

This course had it all!

Paper Reviews #

Every week there were 2-4 research papers assigned to us. For each paper (if you are not presenting it yourself) you were required to write a paper summary in 5-8 sentences. Sounds easy? Well, this was far from the truth! When each paper can range anywhere from 12 to 32 pages long (including appendix) and you go through it completely and understand the content, you feel like writing more than just 5-8 senences. Reflecting back, I think that was the intention of it all along.

For a person, who has stuggled to read research papers, and found it hard to cope with even 1-2 research paper every two weeks in sem 1, this workload felt intimidating! The top 3 questions in my mind at that time were:

• How do I read and understand the papers quickly?
• How do I write paper summaries after reading the paper?
• How do I organize all the papers and the notes for them?

I’ll address everyone of them in their own blog posts.

Paper Presenations #

In the course of the semesters, each students had to present 2 papers in front of the class. For each paper presentation there would be 4 students involved - 2 presentors, 2 respondents. Out of the 2 presentorrs, one would be presenting the pros (10 mins) and the other, the cons (10 mins). After this the 2 respondents would ask questions to the presentors (5 mins).

For e.g. There was a google sheet released at the beginning of the semester with 5 coloumns and students had to sign themselves up for presenting the paper they were most interested in.

Naturally you can see, in a 2 hours class having 4 presentations 30 mins each, there is no time for the prof to “teach” you something in the classical way. Although Prof Lin Shao did spend 10-30 mins at the beginning of some classses touch upon the topic briefly, it was not enough to have in depth knowledge of the topic. Moreover, these intros from the prof were just before students would go up to the podium to present, which means, the class had already spend the previous week reading the papers associated with topic and had submitted their paper summaries the previous night. This means, when you’re learning a new topic for a week, you’re pretty much on your own.

Course Project #

We won the best presentation award.

• Talk about this is another post.

Conclusions #

• My take on this….