Instructor

Nandyala Hemachandra

Semester

Spring 2020

Course Difficulty

High

Time Commitment Required

3 hours of lecture a week, cannot be watched on 2x later as it requires a good amount of understanding and comprehending. Reading around 6-7 research papers.

Grading Policy and Statistics

Definitely not chill, most of the class got a BC

Attendance Policy

Nope

Pre-requisites

Basic knowledge of probability and statistics, a fair bit of ML, RL was expected to be known. It is not an introductory course and Markov chains basics are supposed to be known.

Evaluation Scheme

Quiz 1
Midsem
Endsem
Project

Topics Covered in the Course

Markov Decision models
Single Agent RL
Multi agent RL
SVM
Convex Optimization
Statistical models
Bayes classifiers, Tong Zhang inequality
Decentralized MARL with Networked Agents

Teaching Style

The lectures were conducted on MST with three hours of lecture each week. The prof had ppts with basic points scribbled for him to refer to. These would make absolutely no sense if you just referred to them before the exams without watching the lectures or studying the topic from some other source.

The professor somehow makes the content a lot more tougher by explaining it in a way understood to only 3-4 people in the class. I had to listen to his lectures twice sometimes to understand what was going on. Google and YouTube will be your best friends since you won’t be able to relate one thing to other without it. It’s not that the professor is uninterested, he doesn’t have a flair for teaching, atleast an online class. He expects some basic knowledge on the topics beforehand and if you do have that, it’ll be slightly easier on you.

Tutorials/Assignments/Projects

The practice problems were tough, but those were the kind of questions you could expect in the exams, so you’d be better off if you atleast had a look at it.
The prof had great expectations from the project and will require you to put in a great deal of effort in data collecting, curating, pre processing, cleaning etc. He would appreciate original ideas, but that’s about it, not much help is gained from the professor if you’re stuck. Won’t be like the UG projects which we usually complete (or start, in some cases) very close to the deadline.

Feedback on Exams

Exams were open books, open notes. Some of the questions were directly from the lectures/ papers that he’d provide as reading references. It wasn’t very tough if you go through the material and know where to look for the answers.

Motivation for taking this course

Having done a few introductory courses, the topics seemed interesting

Course Highlights

Nope, can’t think of any, sorry

Course Importance

The topics covered are actually very important and heavy research topics and do have a lot of potential to be explored further, if you do manage to understand them.

How strongly would I recommend this course?

The topics are really good, but the professor’s way of teaching doesn’t let you explore it as best as you could’ve.

When to take this course?

I had taken it up in Sem 6
Ideally to be done in sem 6 or later

Going Forward

Some topics like Multi agent systems, Convex Optimizations etc can be very research intensive

References Used

Additional papers will be provided by the Professor himself, as he covers the topics individually.

Other Remarks

Take up this course only if you’re absolutely sure you can cover the content on your own and can put in a lot of efforts and time into it. The grading isn’t that good and it’s not worth the time for students looking for easy grades. It however does expose you to different topics that would make up individual courses themselves

Review By: Kriti Agarwal