← Quiz: Logistic Regression from Scratch: all 10 questions
Quiz: Logistic Regression from Scratch · Question 10 of 10
Why do libraries clip probabilities to a range like [1e-15, 1 − 1e-15] when computing log loss?
Why do libraries clip probabilities to a range like [1e-15, 1 − 1e-15] when computing log loss?
- A.To make training faster
- B.To change the threshold
- C.To scale the features
- D.To avoid log(0), which is minus infinity
Answer & explanation
Correct answer: To avoid log(0), which is minus infinity
A probability of exactly 0 or 1 would make the loss infinite.
Can you get the rest right without peeking?
Take the timed Quiz: Logistic Regression from Scratch quiz free. Pass it to earn XP and count towards your certificate.
Start the quiz →