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 and Infographic
Here's the deal — b is correct because log-loss = -log(1 - predicted_prob) for a negative-class observation. At predicted_prob = 0.9: loss = -log(0.1) ≈ 2.30.
Full explanation below image
Full Explanation
B is correct because log-loss = -log(1 - predicted_prob) for a negative-class observation. At predicted_prob = 0.9: loss = -log(0.1) ≈ 2.30. At 0.6: -log(0.4) ≈ 0.92. At 0.5: -log(0.5) ≈ 0.69. At 0.3: -log(0.7) ≈ 0.36. The highest confidence wrong prediction (0.9) produces the highest penalty.