LMLearning Monitor

Neural Networks, Visually

Signals Through a Network

8 min · beginner

When a photo app recognizes a face, it does not begin with a ready-made idea of a face. It passes tiny signals through several layers, changing them a little at every step.

Move the slider. Notice how one shared weight scale changes the brightness of the network.

Layers transform signals

The first circles receive inputs. The middle circles are a hidden layer: they transform those inputs. The last circle produces an output. When you moved the weight scale, each connection changed how strongly it carried a signal. Those adjustable connection strengths are called weights.

Check yourself

What does a hidden layer do?
What changed when you moved the slider?