Reinforcement learning, from 0 to something in 60 days
In this series, I will attempt to learn something about reinforcement learning in a limited period of time, after work hours, with the intent to have a somewhat performant system built by the end of that period. A system that plays Atari games maybe, or a trading system — we will see.
There is no intended audience for this series, though if you wish to follow this, you should know that I know some fundamental of machine learnings but am no expert. I will try to learn as much as possible, with 50% planning, 50% improvisation as I will run into yet unknown unknowns. If you are a pro, you might find this boring and inaccurate, if you know nothing about maths and stats, I cannot guarantee you will like this either.
Below is my work plan, which will also be the summary of my findings, when I produce the associated articles. I will come back to this work plan and update it as we go. Today is Feb 28, 2018, and here is what I think we will explore, until end of April 2018:
- A verification that even single neurons can do most of the things I thought was pretty advanced and cool back in university
- A reflection on why neural networks don’t actually use those amazing single neurons and instead use simpler versions of them
- A super fast rundown of Tensforflow basics, and maybe a couple of examples of neural nets created with Tensorflow
- Reinforcement learning fundamentals (part 1, part 2)
- Applied reinforcement learning with Tensorforce, and a working system
