▶Hidden layer
Each hidden neuron applies its own weighted transformation and activation.
Combining Neurons into a Network
Building intermediate representations in hidden layers
A network stacks neurons so that hidden layers build intermediate features and later layers recombine them into an output.
watchone idea
→manipulateone example
→leave withone intuition
The essential idea: composition lets simple neurons cooperate to represent patterns no single neuron can express.
One Concept · One Example
What to notice
- Every layer receives the previous layer's activations.
- Hidden units learn intermediate representations.
- A forward pass applies each transformation in order.
From neurons to a layered map
Output layer
The next layer recombines those hidden activations.
Follow the slides network, edge by edge
Use the one-hidden-layer example from the slides. Nonzero and zero-weight edges reveal exactly how the two inputs become h₁, h₂, and then y.
Hidden neuron 1h₁ = σ(0·2 + 1·3.0) = 0.9526
Hidden neuron 2h₂ = σ(0·2 + 1·3.0) = 0.9526
Output neurony = σ(0·h₁ + 1·h₂) = 0.7216
Hidden layers build reusable intermediate features by composing weighted transformations and activations.