Skip to content

Log Loss gradient and hessian returns NaN for large negative values #28046

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

Jump to bottom
Closed
lorentzenchr opened this issue Jan 2, 2024 · 0 comments · Fixed by #28048
Closed

Log Loss gradient and hessian returns NaN for large negative values #28046

lorentzenchr opened this issue Jan 2, 2024 · 0 comments · Fixed by #28048
Labels
Bug Needs Triage Issue requires triage Numerical Stability

Comments

@lorentzenchr
Copy link
Member

Describe the bug

The private HalfBinomialLoss gradient and hessian returns np.NaN for large negative values of raw_prediction:

  • gradient
  • gradient_hessian
    Only the loss_gradient returns the correct gradient.

Steps/Code to Reproduce

import numpy as np
from sklearn._loss import HalfBinomialLoss

loss = HalfBinomialLoss()
y_true, raw = np.array([1.]), np.array([-1e3])
[
    loss.gradient(y_true, raw),
    loss.loss_gradient(y_true, raw),
    loss.gradient_hessian(y_true, raw),
]

Expected Results

gradient = -1 and hessian = 0

Actual Results

[array([nan]), (array([1000.]), array([-1.])), (array([nan]), array([nan]))]

Versions

sklearn: 1.3.2
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
Bug Needs Triage Issue requires triage Numerical Stability
Projects
None yet
Milestone
No milestone
Development

Successfully merging a pull request may close this issue.

FIX more precise log loss gradient and hessian lorentzenchr/scikit-learn
1 participant
@lorentzenchr