You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Thanks for your interest in WS. Here, the iteration size is basically "iter_size" in caffe, which specifies the number of iterations to take before one parameter update. You can check the implementation of "iter_size" in caffe for more details and why they are equivalent.
Hi authors,
This is an impressing work. However, I am confused about your classification experiments for small batch size.
'setting batch size to 256 and iteration size to 1 (for example) is equivalent to setting batch size to 1 and iteration size to 256.'
Why are they equivalent? For the first case, it only updates parameters 1 times while the second case updates parameters 256 times? Am I right?
Thanks.
The text was updated successfully, but these errors were encountered: