Artificial intelligence and Machine Learning: An introduction

Original article was published on Artificial Intelligence on Medium

Types of Machine learning algorithms

Supervised learning —

Supervised learning uses labeled data to predict something. for example, you can give your program lots of pictures of dogs and label them “dog”, and then give it lots of pictures of cats and label them “cats”. the ML algorithm will then be able to classify images he hasn’t seen yet, as I wrote earlier.

Unsupervised learning —

With unsupervised learning, the algorithm simply finds clusters or patterns in the data. For example, predicting what music someone is more likely to enjoy based on the intensity and the tempo of songs he previously liked or disliked. see an example below:

Reinforcement learning —

Reinforcement learning is very different from what we’ve seen before. when we think about AI we tend to think about reinforcement learning more than about supervised learning or unsupervised learning. Reinforcement learning is based on a reward/punishment system. If the program does something good we give it a reward, else, we give it a punishment. The goal of the program is to maximize its reward. An example of this is self-driving cars. Basically, we can simulate an environment of road traffic and pedestrians and create a car to drive itself in this simulated environment. Every time it crashes or burns a red light, we give him a punishment, if he does good, we give him a reward. of course, self-driving cars are a lot more complex and cannot simply be done in a simulated environment, but this is just an example of how reinforcement learning can be applied. A simpler example is playing a video game such as Flappy Bird, every time the program passes an obstacle, he gets a reward, and if he hits an obstacle, he is punished. A last example is teaching a robot to pick up an apple, if he picks up the apple in a certain amount of time, he gets a reward, else, he is punished.

he is a video of an AI that learns to play hide and seek with reinforcement learning. it uses a little different approach then what we’ve talked about, but everything is explained there, I highly recommend this youtube channel if you are interested in AI (All credits go to the youtube channel: “Two minute papers”):