Week 2/Question 8


in the 8th question of the second week quiz, it was note that “The results from this analysis implies that the team’s highest priority should be to bring more foggy pictures into the training set so as to address the 8.0% of errors in that category.”

I accept that we should expose our system to more knowledge about the foggy states to (via providing more images). but i think we should add these images to dev and test sets; not training set(as it was referred to in the question).

is not it?

I was thinking the same thing. But since dev/test sets are supposed to be our target (best mimic what we will get in real world), we should not temper with them just to reduce error.