Skip to content
ML Academy

Introduced on day 31

Neuron (unit)

One weighted sum of the inputs plus a bias, usually followed by a nonlinearity; a small regression that a layer repeats many times side by side.

In detail

A unit computes z=w⋅x+bz = w \cdot x + b and then an activation such as ReLU or sigmoid. Logistic regression is a single unit with a sigmoid. The biological analogy is historical and weak: treat a unit as a parameterized function, not as a model of a brain cell.