Skip to content

fix: avoid log(0) in KL divergence (Fixes #12233) #13635

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 2 commits into
base: master
Choose a base branch
from
Open

fix: avoid log(0) in KL divergence (Fixes #12233) #13635

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
cf0ce13
fix: avoid log(0) in KL divergence by filtering zero entries
Monasri29-hub Oct 20, 2025
008d408
[pre-commit.ci] auto fixes from pre-commit.com hooks
pre-commit-ci[bot] Oct 20, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
15 changes: 13 additions & 2 deletions machine_learning/loss_functions.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -653,13 +653,24 @@ def kullback_leibler_divergence(y_true: np.ndarray, y_pred: np.ndarray) -> float
>>> predicted_probs = np.array([0.3, 0.3, 0.4, 0.5])
>>> kullback_leibler_divergence(true_labels, predicted_probs)
Traceback (most recent call last):
...
...
ValueError: Input arrays must have the same length.
>>> true_labels = np.array([0.0, 0.3, 0.7])
>>> predicted_probs = np.array([0.1, 0.3, 0.6])
>>> float(kullback_leibler_divergence(true_labels, predicted_probs))
0.10790547587908085
>>> true_labels = np.array([0.0, 0.0, 1.0])
>>> predicted_probs = np.array([0.2, 0.3, 0.5])
>>> float(kullback_leibler_divergence(true_labels, predicted_probs))
0.6931471805599453
"""
if len(y_true) != len(y_pred):
raise ValueError("Input arrays must have the same length.")

kl_loss = y_true * np.log(y_true / y_pred)
# Filter out entries where y_true is 0 to avoid log(0)
# By definition of KL divergence: 0 * log(0/q) = 0
mask = y_true != 0
kl_loss = y_true[mask] * np.log(y_true[mask] / y_pred[mask])
return np.sum(kl_loss)


Expand Down