Hard

Have a go at it. The editor and the docs panel are open, and your code is saved as you type. Running it needs a free account — you’ll come back to exactly what you wrote.

Sigmoid Gradient (Backprop)

~15 mincode completion

Sigmoid Gradient

During backpropagation, we need the derivative of each activation function. For sigmoid:

This elegant formula means: given the forward-pass output , the gradient costs almost nothing to compute.

The sigmoid gradient is maximized at $z = 0$ (value: 0.25) and approaches 0 for large $∣ z ∣$ . This causes the vanishing gradient problem in deep networks, gradients shrink as they flow backward through many sigmoid layers.

Your task:

Implement sigmoid_gradient(z) that returns the derivative of sigmoid at each element of z.

Example Tests

z=0: maximum gradient = 0.25

Input: {"z":0}

Expected: 0.25

Large positive z: gradient near 0

Input: {"z":100}

Expected: 0

Symmetric: gradient same at +1 and -1

Input: {"z":1}

Expected: 0.19661

Python

import numpy as np

def sigmoid_gradient(z):
    """
    Compute the gradient of the sigmoid function.

    Args:
        z: Scalar or NumPy array

    Returns:
        sigma(z) * (1 - sigma(z)), same shape as z.
    """
    # YOUR CODE HERE
    pass

Loading docs…