What are Neural Networks? How do they work? [Complete Explanation]

Today, we live in a world full of Artificial Intelligence. Artificial Intelligence is like giving a brain to the computer. Using AI (Short Form for Artificial Intelligence), computers acquire decision making power based on the previous experiences and data provided by the creator. I won't dive deeper into AI for now You can read more about AI in my other post.

What are Neural Networks?

Artificial Intelligence is based on Neural Networks. It is like the backbone for AI. They are an efficient way of Inputting, processing and outputting data. I will try my level-best to explain neural networks in simple terms.

Basic Structure of a Neural Network

Shown above is a basic bare-bones structure of a Neural Network. A neural network basically has three types of layers:

• Input Layer - It is where the user gives input to the program. Like in a calculator, the place where you enter the numbers and operation.

• Hidden Layers - It is where the real magic happens. It is the layer which is not visible to the user. User cannot interact with this layer. There can be many hidden layers in a neural network. Data is processed and analyzed in different hidden layers simultaneously which reduces the execution time of the program / app. Each hidden layer provides processed data further to other hidden layers for more processing. If an image of a person is provided to the neural network, it would be broken into pieces in these layers. The layers will then match the eyes, nose etc. here. It will then produce result finally.

 

• Output Layer - It is a layer that gives output to the user. The user can see the output printed on the screen due to this layer.

What is a Layer in a Neural Network made up of?

A layer in a neural network is made up of "nodes". Each node is like a single processing unit. Nodes are often referred to as "Neurons". In the diagram above, each circle is represented as a node. Each node processes information and passes that information for further processing to other nodes. Many nodes together make a layer which then combines to form an application.

Now you might be wondering that neural networks are not too complicated and are very easy, but that's not the case. It involves a lot of mathematics (Including calculus too!) and programming. It is very difficult to build a neural network from scratch so we use the pre-trained models available on the internet.

This was just a brief introduction to AI (not even 1%). In the upcoming posts, you will be experiencing what real AI looks like. You need to learn Python in order to continue. A strong knowledge of Python is recommended. You can check out my other posts to learn Python.