Binary Cross-Entropy Loss
MediumMachine LearningMath
Description
Given binary labels yTrue (0 or 1) and predicted probabilities yPred (each strictly between 0 and 1), return the binary cross-entropy loss using natural log: -mean(y·ln(p) + (1-y)·ln(1-p)). Round to 4 decimal places.
Examples
Input:
yTrue = [1,0,1], yPred = [0.9,0.1,0.8]Output:
0.1446Explanation:
Averaging the negative log-likelihood across the samples gives a loss of 0.1446.
Input:
yTrue = [1,1], yPred = [0.5,0.5]Output:
0.6931Explanation:
Averaging the negative log-likelihood across the samples gives a loss of 0.6931.
Input:
yTrue = [0,1], yPred = [0.2,0.7]Output:
0.2899Explanation:
Averaging the negative log-likelihood across the samples gives a loss of 0.2899.
Constraints
- •
1 ≤ yTrue.length ≤ 10⁴ - •
0 < yPred[i] < 1