Applications of Reinforcement Learning: In Games
In 2016, AlphaGo became the first program to win at the game of Go against a professional player. Google’s AlphaGo defeated world-renowned player Lee Sedol four matches out of five. Go is a very demanding game requiring strategic and adaptative thinking. Yes, it’s even harder than chess! It is astonishing that a computer was able to surpass an extremely skillful human.
You may be asking yourself who taught AlphaGo to be this good. Believe it or not, the computer taught itself! This is all thanks to reinforcement learning. After being given millions of Go matches to analyze, AlphaGo was able to learn by itself: by being rewarded for taking the right decisions and punished for taking the wrong ones, the machine was able to establish a system to maximize the reward probabilities, hence making the best moves.
