Two numbers go in. One number comes out. Watch what happens in between.
One neuron. The two blue circles on the left are its inputs. The purple wires carry a weight each — thicker means it counts for more, red means it counts against. The circle in the middle adds everything up. The gate on the right is the activation, and the teal circle is what the neuron finally outputs.
Press Send a signal and follow one value through. Then change the inputs, the weights or the bias and send another.
Each input is multiplied by its weight. Those products are added together, plus the bias — a number the neuron carries on its own. That total is called z, the pre-activation. The activation then turns z into the output: a ReLU passes z through when it is positive and outputs 0 when it is negative.
That is the entire neuron — multiply, add, then one simple decision. Weights decide which inputs matter and in which direction; the bias shifts the point where the neuron switches on. Everything in a deep network is billions of this.