#0020

Binary Cross Entropy From Probabilities

MediumML18 min15 XP

Problem

Given true labels y_true and predicted probabilities y_prob, return average binary cross entropy. Clip probabilities to avoid log(0).

Why This Matters

Binary cross entropy is the loss behind many churn, fraud, medical-risk, and click prediction models.

Function Signature

def binary_cross_entropy(y_true, y_prob):

Examples

Example 1

Inputy_true = [1, 0], y_prob = [0.9, 0.2]

Outputabout 0.164

Confident correct probabilities produce a low loss.

Constraints

y_true contains 0 and 1 labels.
y_prob contains predicted probabilities.
Return the mean loss as a float.

CodePython

import numpy as np

def binary_cross_entropy(y_true, y_prob):
    """
    Input example: y_true = [1, 0], y_prob = [0.9, 0.2]
    Return: about 0.164
    Return the value from the function so the tests can compare it. Do not print unless the prompt asks.
    """
    # write your code here
    return 0.0

import numpy as np

def binary_cross_entropy(y_true, y_prob):
    """
    Input example: y_true = [1, 0], y_prob = [0.9, 0.2]
    Return: about 0.164
    Return the value from the function so the tests can compare it. Do not print unless the prompt asks.
    """
    # write your code here
    return 0.0

Visible browser tests run here when available.

Testcases2 visible / 4 hidden categories

Confident mostly correct predictions

Input[1, 0], [0.9, 0.2]

Expectedabout 0.164

Both predictions lean toward the correct class.

Avoid log zero

Input[1, 0], [1.0, 0.0]

Expectedfinite loss near zero

Clipping prevents log(0) from producing infinity.

Hidden Test Categories

All positive labelsAll negative labelsBad confident predictionsProbabilities close to 0 or 1