Machine learning, in a broad sense, is the process of making machines learn from a set of data, and thus, their performance on a specific task improves without being the type of programming. It’s like teaching a child how to sort different shapes: at the beginning, you give directions, like driving circles, triangles, and squares, which show in different colors. You want to introduce the child to the knowledge of recognizing and correctly grouping these shapes into their categories and carrying out this task by taking the steps below.

1. Gathering data

First, you give a set of choices with shapes and define the group to which each of the presented shapes belongs. You familiarize the child with some of the colors and shapes that they may be able to recognize. For instance, you could teach a circle with green color and give a narration like, “This circle is green”. Next, the same is done to the children to show them the various colors and shapes in the same manner as described above.

2. Training a model

Now, the child may see even more examples of shapes and colors but without an idea of what exactly to do with them. Construct a game provisioning a picture of a shape that easily guesses which category it belongs to. This can actively take place in the process of every thought, where you inform that the answer is either right or wrong.

3. Predictions

With the play in progress, the child will see more patterns, and the answers will be guessed more precisely. To give a more concrete example, a young child would see a yellow triangle and possibly guess that the shapes are yellow and that it would be a triangle. Bit by bit, they become good at it to the point where they finally make a correct guess with the help of their experience and errors.

4. Feedback loop

With every guess, the child will learn clearly, and facts will be underlined by explaining why this answer is incorrect. These comments will allow the child to learn, which will allow them to make an attempt to get it right.

The purpose of the above post is to explain machine learning using the child-teaching example. Much like a child learns about shapes by sorting them using examples and getting feedback, machine learning algorithms get better with time as they learn from data and feedback. There are three categories of machine learning: supervised, unsupervised, and reinforcement learning.