What is log loss (cross-entropy) and what does a specific value mean?
mediumAnswer
- For binary: log loss = -1/n * sum [y log(p) + (1-y) log(1-p)].
- It's the negative log-likelihood of the observed labels under the predicted probabilities — proper scoring rule that rewards well-calibrated confident predictions and heavily punishes confident mistakes. ≈ 0.693 is the loss of always predicting 0.5 (max-uncertainty baseline).
- Above that, your model is worse than uninformative; below, it's actually learning something.
- Never use log loss for evaluation on a mis-calibrated model.
Check yourself — multiple choice
- Log loss ignores predicted probabilities
- It's the negative log-likelihood; ≈ 0.693 is the uninformative baseline
- Log loss can be negative
- It only works for regression
Log loss = NLL. 0.693 ≈ always-0.5 baseline. Proper scoring rule.
#metrics#classification#calibration
Practise Supervised Learning
215 interview questions in this topic.
Related questions
- What is the Brier score and how does it compare to log loss?
- What is Youden's J statistic and how is it used?
- How do you select the decision threshold for a binary classifier?
- Precision vs recall — when do you optimize each?
- How do you read a confusion matrix and derive the key metrics?
- How does temperature scaling calibrate a classifier's probabilities?