Types of Machine Learning

Supervised Learning

The most popular supervised learning tasks are:

• Regression (for example, predicting the price of a house): you will need a training set labeled with other house prices for that;

• Classification (for example, classifying email as spam/ham): you will need a training set labeled as spam/ham for that.

Unsupervised Learning

The most popular unsupervised learning tasks are clusterizaion, anomaly detection, and dimensionality reduction.

Clusterization

It is a process of grouping similar data points into clusters. You do not need to label the data for it. For example, a training set of emails without labels spam/ham will do.

Anomaly Detection

It is a process of detecting deviations from normal data behavior. For example, fraud detection in credit card transactions. No need to label fraud/not fraud. Simply give the transaction information to a model, which will determine if the transaction stands out.

Dimensionality Reduction

It is a process of reducing the number of dimensions while retaining as much relevant information as possible. It also does not require any labels.

Reinforcement Learning

Reinforcement learning differs significantly from the previous two types. It is a technique used to train self-driving vehicles, robots, AI in gaming, and more.

Imagine training a dog to fetch a ball. The dog would receive a reward (such as a treat or praise) for picking up the ball and bringing it closer to the owner. It would receive a penalty (such as withholding the treat or a disappointed tone) if it runs in the wrong direction or gets distracted. Additionally, it would get a large reward once it successfully retrieves the ball and delivers it to the owner.