Dataset and label checks · Question 40 of 60
What can a low prediction confidence of a trained model for a data point mean?
3 answers are correct. Medium
Show answer
Answer
- A The data point is ambiguous
- B The data point is mislabeled
- D The data point lies outside the training data distribution
Explanation
If the model is very unsure about a data point, there can be several reasons: the label is wrong, the case is ambiguous, or it lies far away from anything seen in training. Such data points are therefore examined more closely.