A number of us from the lab have been helping to put this course together. It's lead by Martha White and Adam White - two awesome RL profs at the U of A (Martha now leads RLAI) - and is based very heavily on Rich's textbook. The goal is to provide a really strong foundation for those looking to dive deeper into reinforcement learning. It starts with bandits and works all the way up through function approximation, control, policy gradients, and deep RL.
If you have any questions feel free to ask and I'll do my best to answer.
> there's still no great resource to learn RL "from scratch" - there's still a huge gap between Sutton&Barto and implementing DDPG. You have to figure out everything by reading existing implementations, various Medium posts (a lot of them containing errors and imprecisions), and research papers. I wouldn't consider Spinning Up as a beginner-friendly resource, it's too dense/math-heavy. The closest I have found so far is the Udacity course: https://eu.udacity.com/course/deep-reinforcement-learning-na.... which costs $1000
I too think OpenAI's Spinning Up isn't beginner-friendly. But I also don't want to just learn bandits and tic-tac-toe. Will this course fill the gap?
Agreed! A lot of material out there is like the "how to draw and owl" meme: https://imgur.com/gallery/RadSf - start with bandits and now do DDPG.
The goal is for this course to provide the foundations for whatever folks want to do in RL after. It starts with bandits but then covers things like TD, Sarsa, Dyna, etc. in the tabular setting. Then folks learn about more advanced topics like linear and non-linear function approximation (read - linear e.g. Tile Coding, non-linear e.g. neural nets/deep rl).
This very much follows the intro RL course taught by Martha/Adam/Rich at the U of A, and follows Rich's textbook really closely.
Reinforcement learning has been used in a lot of valuable applications to society/the economy outside of games: control system optimization, robotics, ad targeting, content personalization to name a few. Game playing can often be a great test-bed for RL algorithms that can be applied in other areas.
Csaba's book is the most up-to-date on RL I know of. Sutton and Barto is very old by now. For the POMDP side of things there are no recent books I know of, but http://www.cs.mcgill.ca/~jpineau/files/sross-jair08.pdf is a recent enough survey.
The bandit problem is very strongly related to the reinforcement learning problem, so you'll get some mileage out of studying bandits. Be aware this area is very maths heavy, which is good or bad depending on your background. If you like you like this stuff, also checkout "Prediction, Learning, and Games" which deals more with the "adversarial" setup.
I found that Real Python's blog and ebooks were a good way to look at a few different frameworks (especially Flask and Django). There are a few cases where I wish they had explained things further in the web books rather than just show the code, but overall they should be a good next step after LPTHW.
http://www.realpython.com/
What is the name of your Uncle's company? Tell him to fire me an email, I'd be happy to do what I can to help him find someone. There are quite a few good devs in Edmonton and specific to gaming BioWare attracts a large number of talented people to town, and leaks out some very good people at times depending on their development cycle.
cam.linke@gmail.com
I love being able to assign hot keys to apps in Alfred. I have 4-5 apps that I use 99% of the time and now I have hot keys to easily bring them up and jump between them.
If you have any questions feel free to ask and I'll do my best to answer.