What is a feedforward neural network?

A feedforward neural network is a neural network in which information moves in one direction—from the input layer, through any hidden layers, to the output layer.

There are no feedback loops or recurrent connections that send information back to earlier layers.

Basic structure

A typical feedforward network looks like:

Input layer → Hidden layer(s) → Output layer

Each layer receives values from the previous layer, transforms them, and passes the result forward.

What happens inside each layer?

A neuron usually computes a weighted sum:

z = Wx + b

and then applies an activation function:

a = f(z)

The output of one layer becomes the input to the next layer.

Why is it called feedforward?

It is called feedforward because data only moves forward through the network during the forward pass.

Unlike an RNN, a standard feedforward network does not maintain a hidden state from previous inputs.

What are feedforward neural networks used for?

They are commonly used for:

• Classification
• Regression
• Tabular data
• Function approximation
• Final prediction layers inside larger neural networks

How is an MLP related?

A multilayer perceptron (MLP) is one of the most common types of feedforward neural network.

It usually contains one or more fully connected hidden layers with nonlinear activation functions.

For more detail, see “What is a multilayer perceptron (MLP)?”.

In short, a feedforward neural network processes information from input to output in one direction without recurrent feedback connections.