sigmoid

→Tensor

sigmoid(input: Tensor)

Element-wise logistic sigmoid.

Squashes the real line into $(0, 1)$ . Often used for binary classification heads.

inputTensor

Input tensor.

Tensor

Element-wise result with the same shape as input.

Mathematical definition:

\text{out}_i = \sigma(x_i) = \frac{1}{1 + e^{-x_i}}

Gradient: $\sigma(x)(1 - \sigma(x))$ . Numerically stable for large negative x via the engine implementation.

>>> import lucid
>>> x = lucid.tensor([1.0, 2.0, 3.0])
>>> lucid.sigmoid(x)
Tensor([...])