Cross-Entropy Loss Calculator
Use this calculator to evaluate the cross-entropy loss between your true labels (one-hot) and model predicted probabilities for classification problems.
What this calculator is doing
This calculator computes the cross-entropy loss, a standard loss function in classification tasks, using:
\[
\text{Loss} = - \sum_{i=1}^{n} y_i \cdot \log_b(p_i)
\]
Where:
- \( y_i \): True label (1 or 0, from one-hot vector)
- \( p_i \): Predicted probability for class \( i \)
- \( b \): Logarithm base (e = natural log, 2 = bits, 10 = common log)
Only the non-zero \( y_i \) values are included, meaning typically only one term in the sum.
This metric is used to measure the dissimilarity between the true distribution (label) and the predicted
distribution, where lower values indicate better predictions.
Choose from base e, 2, or 10 depending on your application (e.g., entropy in nats, bits, or bans).
Disclaimer: These calculators are provided for informational purposes only. Always verify your designs against relevant engineering standards and consult a qualified professional. We do not take responsibility for any errors or damages resulting from the use of these calculations.