What is Machine Learning (ML)?

Machines as we know, are good at computation. In terms of computation, there are 3 elements involved:

  • Input
  • Function
  • Output

Typically, computation applies a given function on given input, and return some kind of output. However in Machine Learning, Input and Output are given, and the job of the machine is to figure out the Function which describes the relationship between them, also known as model, or h for hypothesis

Are all Machine Learning the same?

NO. There are mainly 2 types of Machine Learning:

  • Supervised-learning
  • Unsupervised-learning

More on supervised-learning

In supervised-learning, input and output are provided. The job is to figure out the relationship between them. Supervised-learning can be futher categorized into Regression problem and Classification problem.

Regression

When the output is a continous number. Eg:

  • predicting price of the house
  • predicting a person’s income by certain age

Classification

when the output is discrete, we can think of them as categories. Eg:

  • classify the spam emails
  • determine whether a person has diabetes

Sometimes, the result of regression problem can be used to solve a classification problem. Eg, given the predicted price of the house, whether one could sell it successfully. That’s also known as Logistic Regression

More on unsupervised-learning

In unsupervised-learning, only a set of inputs are given. There is no output. The goal is usually finding clusters or groups that are closely related given certain inputs. Eg:

  • given 10000 articles, put them into groups with similar topics