What is log loss?

Updated May 15, 2026

Short answer

Log loss measures performance of classification models using probabilities.

Deep explanation

It penalizes wrong predictions based on confidence, derived from likelihood functions in statistics.

Real-world example

Used in logistic regression for binary classification.

Common mistakes

  • Using raw predictions instead of probabilities.

Follow-up questions

  • Why use log in loss?
  • What is relation to likelihood?

More Cost Function interview questions

View all →