How a neural network moves information
Goal: understand what a neuron receives, what a connection changes, and how values move forward through layers. This lesson stops at forward propagation; training and backpropagation come later.
Watch the visual introduction
Use 3Blue1Brown's handwritten-digit example through 13:32. Stop before the compact vector and matrix notation begins.
Open “But what is a neural network?” ↗Walk through five networks
Open the playground and change presets in this order:
- 1 → 1Follow one input across one weighted connection.
- 2 → 1See how a neuron combines more than one input.
- 1 → 2 → 1Introduce a hidden layer and two intermediate values.
- 2 → 2 → 1Compare several weighted paths through the same layer.
- 3 → 3 → 2Finish with multiple inputs, hidden neurons, and outputs.
Let members change one thing
Select an input, connection, or node. Change its value, weight, or bias and observe which later values move. Keep the activation on ReLU: negative becomes 0, while a positive value stays unchanged.
Resources inside this lesson