What is it?
Like a lot of the ideas in machine learning, it is an abstraction of real life. How do we learn in real life? By doing and seeing the results. Put your hand in a fire and you won't do it again. You just learned something: fire is hot.
That lesson is taught by negative reinforcement. Having your hand burnt is negative it doesn't feel good. Alternately you can reward positive actions. Like a baby trying ice cream.
Setting the scene
The first step in reinforcement learning is establishing an environment. This is distinct from a programming environment. It can be something as simple as an empty array that represents a game board or you could have a robot learning in the real world.
Top comments (0)