Reinforcement Learning: AWS DeepRacer

I have no intention of writing yet another article on Reinforcement Learning; you will find plenty of them from true experts.

I just want to share my experience of getting familiar with Reinforcement Learning through AWS DeepRacer. I have divided my writing into two parts.

• Quick Introduction to Reinforcement Learning (RL)
• Using RL through AWS DeepRacer

Reinforcement Learning

This is a diagram Ande Jessey showed to explain RL in re:Invent 2019. It is pretty self-explained where RL comes into play. In supervised learning, along with sophisticated models, you need a good amount of labeled / training data; for Reinforcement Learning, you need very little training data.

So you can see, where supervised learning is more about “curve fitting”, reinforcement learning is all about getting rewarded for your progress.