Log-loss penalizes confident wrong predictions much more heavily than uncertain wrong predictions. Which scenario produces the highest log-loss for a single observation?
Select an answer to reveal the explanation.
Short Explanation
Here's the deal — log-loss punishes confident wrong guesses hard. For a negative-class observation, the loss is -log(1 - predicted probability), so a 0.9 prediction gives -log(0.1) ≈ 2.30, the biggest penalty.
Full Explanation
For a negative-class observation, log-loss is -log(1 - predicted probability). The 0.9 negative-class prediction gives -log(0.1) ≈ 2.30, the highest penalty. The 0.6 negative-class prediction gives -log(0.4) ≈ 0.92; the 0.5 negative-class prediction gives -log(0.5) ≈ 0.69; the 0.3 negative-class prediction gives -log(0.7) ≈ 0.36. The most confident wrong prediction is penalized most.