Hidden Layers

Where Are Hidden Layers?

A simple neural network can be viewed like this:

Hidden layers are called "hidden" because they are not directly visible as the input or final output of the model.

Why Do We Need Hidden Layers?

Real-world problems are often more complicated than simply connecting an input directly to an answer.

Hidden layers allow the network to process the input in multiple stages and learn more complex patterns.

Simple Example — Student Prediction

Suppose we want to predict whether a student will pass an exam.

The input might contain:

The input layer receives these values.

The hidden layer then processes those values and can learn relationships between them.

The important point is that the hidden layer does not simply copy the input. It transforms and processes the information before it reaches the output layer.

Can a Network Have More Than One Hidden Layer?

Yes. A neural network can contain multiple hidden layers.

Each hidden layer receives information from the previous layer and passes processed information to the next layer.

Hidden Layers Learn Patterns Step by Step

A useful way to understand hidden layers is to think about learning patterns in stages.

For example, imagine a neural network processing an image of a face.

The exact behavior depends on the architecture and training, but this is a useful beginner-friendly way to understand why multiple layers can help a network learn increasingly complex representations.

Hidden Layers Contain Neurons

A hidden layer is not a single neuron. It normally contains multiple neurons.

Each neuron performs calculations on information received from the previous layer.

We will study exactly how a neuron performs those calculations later in this lesson.

How Information Moves Through Hidden Layers

Information flows from one layer to the next.

Each layer transforms the information before passing it forward.

How Many Hidden Layers Do We Need?

There is no single number that works for every problem.

A simple problem might work with a small network, while a complicated problem may require a much deeper architecture.

Why Is This Called Deep Learning?

The word "deep" generally refers to using multiple layers in a neural network.

More hidden layers can allow a network to build increasingly complex representations from the input.

Hidden Layer vs Hidden Neuron

For example, a hidden layer could contain:

Hidden Layer
    ↓
    Neuron 1
    Neuron 2
    Neuron 3
    Neuron 4

Complete Neural Network Flow

So the overall idea is:

Input
  ↓
Hidden Layer
  ↓
Hidden Layer
  ↓
Output