I've heard of reinforcement learning in a webinar and have always been curious about it so I decided to read more about it. Here's what I've learnt:
Reinforcement learning is one way to train a machine learning model and achieve a sequence of decisions. It is very useful when teaching robots how to walk or mimic human-like actions. Here you train the agent (the humanoid/the object learning to make decisions) in an uncertain and complex environment. So in the example of walking, you would create obstacles like walls or speed bumps along the path for the agent to recognize and avoid. This is where the computer while trying a trial-error approach and come up with a solution to the problem.
To get the machine to actually do what the programmer wants, you award rewards for current actions or penalties for the actions you don't want it to perform. The computer's end goal is to maximize the total reward points.
Essentially this is like teaching good behaviour to a small child and using rewards to reinforce the teachings.
Log in or sign up for Devpost to join the conversation.