How does a neural network work?
A neural network turns input into output through layers of connected computing units.
What is it?
A neural network is a mathematical model made of layers of connected computing units. Each unit adds up its inputs, weights them and passes the result on. During training, the network adjusts the weights so its outputs come as close as possible to the right ones.
An everyday example
When your phone camera recognises a face and focuses on it, a neural network works out step by step from the pixels where the face is.
Why does it matter?
Neural networks are the basis of modern AI, from translators to chatbots. Understanding how they work makes their limits easier to understand too.
A common misconception
That an artificial neuron works like a nerve cell. The name is a useful metaphor, not a description of the biological brain.
In more detail
What are artificial neurons?
Simple computing units that combine numerical inputs and apply a mathematical function.
What does the network learn?
The values of its connections and other parameters that help it produce better outputs.
What is backpropagation?
A calculation that works out how changing each parameter would affect the error.
Is there just one type of network?
No. Different designs suit images, sound, text, time series and the physical world.
Remember: The name “neuron” is a useful analogy, not a description of a real nerve cell.
Reliable sources
Last reviewed 1 November 2026. An explainer, not professional, legal or medical advice. For rules and figures that change, check the official version.