Layers Working Together

Think about making a sandwich. First you lay out the bread, then you add cheese, then lettuce, then tomato, and finally the top slice of bread. Each layer does its own job, and together they make something delicious. A neural network is organized in layers too — and each layer has its own job to do.

Three Kinds of Layers

A neural network has three types of layers, and every network has at least one of each. The first type is the input layer. This is where information enters the network. If the network is looking at a photo, the input layer receives all the numbers that describe the photo. The second type is called hidden layers. These are the layers in the middle — the ones doing most of the hard thinking. Hidden layers look for patterns, like edges in a photo, or certain shapes. The third type is the output layer. This is the last layer. It gives us the final answer. For a photo of a cat, the output layer might say: 'Cat: 95% sure. Dog: 3% sure. Bird: 2% sure.'

Here is a real-world story. Imagine the network is trying to figure out if a shape is a square or a circle. The input layer receives the numbers describing the shape — brightness of each dot in the picture. The hidden layers look for clues. One layer might notice straight lines. Another might look for corners. The output layer uses all those clues and says: 'Square!' Each layer passed what it found to the next, and together they solved the puzzle.

When a network has many hidden layers stacked on top of each other, scientists call it a deep neural network. That is where the word deep learning comes from! The deeper the network, the more layers of pattern-finding it can do. Very deep networks can recognize faces, understand speech, and even play video games.