Hidden layers

The network from the last chapter already had a middle step.

Colour and size went into two neurons. One answered yes when a mushroom belonged to the small, pale safe region. The other answered yes when it belonged to the large, dark safe region.

The final neuron answered safe when either middle neuron said yes.

Those two neurons in the middle form a hidden layer.

Hidden only means that the layer sits between the original inputs and the final answer. Its results are steps along the way.

Nothing new has been added. The middle step now has a name.

Every connection into or out of the hidden layer has a weight. If the final answer is wrong, training must trace the mistake back through those weights.

Hidden layers

The network from the last chapter already had a middle step.

Colour and size went into two neurons. One answered yes when a mushroom belonged to the small, pale safe region. The other answered yes when it belonged to the large, dark safe region.

The final neuron answered safe when either middle neuron said yes.

Those two neurons in the middle form a hidden layer.

Hidden only means that the layer sits between the original inputs and the final answer. Its results are steps along the way.

Nothing new has been added. The middle step now has a name.

Every connection into or out of the hidden layer has a weight. If the final answer is wrong, training must trace the mistake back through those weights.