Suggestions to learn Reinforcement Learning

Hi,

Can anyone please point to some good place to start for the intuition development for Reinforcement learning. Is the ML course coming up going to include RL?

1 Like

Hi Bilal

Nice to see that you’re interested in getting into RL. A good first step to building a foundational understanding on the topics would be to go through a well designed online course, you can check this link for some options.

The upcoming ML course part of the PadhAI One series will largely be covering classical Machine Learning techniques and their theoretical + practical implementation. At this point, there is no plan to include RL in the syllabus.

3 Likes

Thanks @Manick a lot
:slightly_smiling_face:

NPTEL also has a 12 week course on Reinformcement learning starting on July 20th. Link below

Reinforcement Learning

1 Like

Just went through the same, thanks @sanjayk.
:slightly_smiling_face:

In the last few days, I have gone through a few online resources related to Reinforcement Learning. My first impressions, just in case somebody needs it: .

  • Udacity has a free course on Reinforcement Learning(Click)
    Pros: The way they teach is pure Joy. Its fun and addictive. One of the course instructor is Michael Littman who is credited with Reward Hypothesis by Richard Sutton(Father of RL). Reward Hypothesis is core on which everything is build in RL.
    Con: They stay on the surface most of the time (don’t go too deep). Having said that, they teach you the concepts in a logical way (and fun way). Advanced course in Deep RL is paid (and probably expensive, but you get udacity’s nanodegree).
  • Coursera has a RL Specialization (Click)
    Pros: Its very organised, slow pace, audit is free(you have access to all study material, videos). You get specialisation certification (paid). after completing a series of 4 courses. Conducted by University of Alberta which is like the home of RL. Instructors (Whites) are basically some of the top names in RL. Considered to be one of the best resources for learning RL. And yes, programming exercises, quizzes etc.
    Cons: I found the teaching uninspiring. You are advised to read from Richard Suttons RL book before watching videos lectures. Lectures are mostly a repetition of reading and doesn’t involve any real addon or explaining of concepts. It feels like a big reading group for the book.
  • David Silver’s youtube lectures: David Silver’s lectures (Click):
    ProsMy favourite so far. I’ve seen the first two lectures. He helps build the concepts gradually. Ideas and building blocks precedes any complex concept(Similar to how Mitesh Sir does in his Deep learning leactures). Takes plenty of examples which helps the learning. Its intense.
    Cons:. Its just youtube videos, no quizzes or programming exercises as you will get with a formal course like the one on coursera or udacity. Lectures are long, atleast an hour and 30 min long, and they are intense :stuck_out_tongue:
2 Likes