Hearts or circles: what the heck is a neural network? Today’s artificial intelligence is built on neural networks. Here you will learn what they are by training a real one, in your browser, in about 30 minutes and with no prior knowledge.

You will draw a circle and a heart, the network will train on hundreds of examples, and you will be able to test it and correct it. Along the way you will see what a neuron is (a judge who adds up evidence of different weight and compares it with a threshold), how neurons form a network, how it learns by trial and error with gradient descent, why memorising is not learning and why a network cannot explain its decisions.

This application needs JavaScript. Please enable it to start. Author: José Carlos Rueda Álvarez (jcrueda.com).

ES

Hearts or circles: what the heck is a neural network?

A hand draws a circle and a heart in chalk on a blackboard; a small network of nodes grows out of the drawing — Work by José Carlos Rueda Álvarez, lawyer · 2026
Text and code under CC BY-ND 4.0; the images, generated with AI, are not covered by the licence This page does not collect or send any personal data: no cookies, no analytics. It only remembers, in your own browser, whether you prefer the light or dark theme.

Board15 × 15 squares

○ circleheart ♥

Prediction without training: the network has not learnt anything yet

What pushes

Inside the networkpoint at a neuron · click to switch it off

How it sees the exampleshidden layer · PCA

What each neuron looks forclick to switch off or on

Why it decides thatsum of contributions

Learning

Examples