English Language Learners Stack Exchange is a question and answer site for speakers of other languages learning English. It only takes a minute to sign up.
Sign up to join this community
I was reading an article about deep learning and I've encountered this sentence:
Usually, deep learning models use dropout on the fully connected layers, but is also possible to use dropout after the max-pooling layers, creating image noise augmentation.
I don't know the meaning of creating image noise augmentation part. Is it the sequence of using dropout after the max-pooling layers or is it going to come alongside it?
It's not clear from the way it's written, but I believe the author intends to say that the image noise augmentation is created as a result of using dropout on the max pooling layers.
Even on their own, max pooling layers can accentuate noise because they select the highest value. By design, dropout introduces noise by eliminating features that might help create a more accurate prediction. So the combination of the two would be more likely to incorporate noise into the final model.
"creating image noise augmentation" is the result of the process, though possibly an un-intended consequence.
What is also not entirely clear, is whether "creating image noise augmentation" would be a result of both options, or (more likely) only a result of the latter option.
And for completeness, but is also possible to use ... should read: but _it_ is also possible to use ...
