Sigmoid

S-shaped activation that squashes any real value into the range \((0, 1)\).

$$ \sigma(x) = \frac{1}{1 + e^{-x}} $$

• Output is a smooth probability-like score
• Used in binary classification output layers
• Gradient vanishes for very large or very small inputs (vanishing gradient problem)
• Computationally more expensive than ReLU due to the exponential