Sigmoid
In plain terms
An activation function squashing values into the 0-to-1 range.
Going deeper
The sigmoid function maps any input to a value between 0 and 1, useful for probabilities, but prone to vanishing gradients in deep nets. It produces a smooth S-curve.
Under the hood
Saturates at extremes (gradients near zero), so it's mostly used in output layers now, not hidden layers.
Examples
Binary classification output, logistic regression, gating, probability mapping.
Think of it like...
Like a smooth dial that eases any input into a 0-to-1 probability.