Bounded, continuous, and always differentiable—Sigmoid presses every signal into the familiar: probability between zero and one. Once the world’s default gate, now a specialist—best for binary classification, logistic regression, and any task where output must be interpreted as a probability. Watch the S-curve: slow at the edges, sharp in the middle. Vanishing gradients lurk for deep networks, but for the right job, there’s no substitute. For those who respect elegant boundaries.
