neuron (artificial)

A unit that multiplies its inputs by weights, adds a bias, sums, and applies a nonlinearity.

An artificial neuron computes $z = \sum_i w_i x_i + b$ from its inputs $x_i$ , weights $w_i$ , and bias $b$ , then passes $z$ through a nonlinear activation function $\varphi$ to produce its output $a = \varphi(z)$ . It is the single computational unit a neural network stacks by the thousands or millions.

Examples

A neuron with weights $(0.5,-0.2)$ , bias $0.1$ , and a ReLU activation, given inputs $(2,3)$ , computes $z=0.5(2)+(-0.2)(3)+0.1=0.5$ and outputs $a=\text{ReLU}(0.5)=0.5$ .