First of all, I want to thank Jakarta Machine Learning and AWS for allowing me the opportunity to join the AWS DeepRacer boot camp. I definitely will share my learning experience during this boot camp through my articles. So, please stay tuned to know and learn more about my boot camp experience!

In order to race autonomously, AWS DeepRacer needs to learn how to drive itself, just like a professional racer drives his car. This learning mechanism is also similar to a toddler that learns how to walk. Similar to the previous example, AWS DeepRacer also applies this method of learning, which is called reinforcement learning.

What is Reinforcement Learning?

At this point, you may be wondering what reinforcement learning means. To put it simply, reinforcement learning is learning by interacting with the environment. It is active and sequential, which means the future depends on earlier interactions. In addition to that, it directs towards a goal and the system can learn without examples of optimal behavior (trial and error).

For an autonomous car, you might ask the following question.

Why can’t we program the car to do exactly where to turn left and right?

Well, there are two reasons to learn. Firstly, we want to find previously unknown solutions. For example, a program that can beat a human chess master. Secondly, we want to find solutions for unforeseen circumstances. For instance, an autonomous car that can navigate tracks that differ greatly from any known tracks.

In other words, reinforcement learning is the science of learning to make decisions from interaction with the environment. This concept is applied in many fields, ranging from computer science to economics.

#reinforcement-learning #machine-learning #aws #rewards #data-science