0
$\begingroup$

I wonder what is considered a better practice in deep learning. I have a dataset of 100K images that I want to use for training a regional-CNN for a whole week.

Is it possible that my network will be better trained if I shrink the dataset, throwing away 75% of the samples (suppose that the data is shuffled and that every class has equal number of samples), just because it will run 4 times more epochs on it?

$\endgroup$
0
$\begingroup$

You would most likely (over-)fit your training data "better" and generalize worse. You may want to look into better algorithms, parallelization and renting compute on the cloud.

$\endgroup$
  • $\begingroup$ Thanks for the answer. In my settings, would you prefer 20 epochs on 100k images rather than 80 epochs on 25k images? Will the "generalization" with 20 epochs be worthy? $\endgroup$ – SomethingSomething Feb 5 at 12:52

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.