Coursera Machine Learning Study notes (iii)

-Unsupervised learning

In supervised learning, whether it is a regression problem or a classification problem, we use the data to have a clear label or the corresponding prediction results.

In unsupervised learning, our existing data have no corresponding results or labels, and some are just features. Therefore, the problem to be solved by unsupervised learning is to find out whether the data can be divided into different groups.

A typical problem of unsupervised learning is clustering problems, such as:

-Analyze data transmission in large data center networks and analyze which machines are working together.

-Social network analysis

-Market Segmentation

-Astronomical Data analysis

-Human gene cluster analysis

Another classic example is the cocktail party problem, where people talk to each other in a room full of people, we use a number of microphones to record the sound of the room, and use unsupervised learning algorithms to separate what one person in the room says.

A simplified version of the cocktail party problem is a room in which two people are speaking at the same time, using two microphones to record.

Coursera Machine Learning Study notes (iii)