This course will take you from complete beginner to reinforcement-learing aficionado over 5 lessons. Each file comes with explanatory comments and exercises to help you develop your understanding of reinforcement learning!
By our estimates, you could take:
- Roughly 1 hour to run and comprehend each of the files provided;
- Roughly 1 day to fully read and comprehend each of the files;
- Roughly 2 days to complete all the basic exercises in each file;
- Roughly 5 days, if you're a programming expert, to complete all the challenge exercises in each file.
Reinforcement learning is a type of machine learning (as opposed to prediction/classification) that allows an agent (e.g. a robot) to learn to respond to an environment. If you've taken non-reinforcement learning machine learning courses before, then reinforcement learning is different because it deals with actions rather than dealing with static data. This makes reinforcement learning particularly relevant to robotics, as it allows our robots to interact with their environment as opposed to just classifying or making predictions about things in the environment.
To get through this course smoothly, you should be confident with python. If you want a refresher, we've got a tutorial for you over here: https://github.com/usydroboticsclub/python
You would also benefit from:
- Some experience with abstract problem solving
- The ability to read code and understand it, regardless of how much/little commenting there is
- A conceptual understanding of neural networks
But give it a go! We trust in your ability to learn :)
Jay Zhang found this excellent Deep RL course available for free from UC Berkeley: http://rail.eecs.berkeley.edu/deeprlcourse/
It's a natural follow on from the current course, and is UC Berkeley standard; and completely free!
In case they've taken it down because it's too good, we'll hopefully have kept a cached version of the slides and exercises somewhere, if Steven ever gets around to it.
Please send feedback to [email protected], or if you want to contact the authors directly, you can reach out to [email protected]
Happy learning!