I have a data set which I have various aspects of the data, so I think a CNN will work best. I have grouped each aspect of the data into sets of 5 data points, a couple of which I have padded with zeroes. My first layer has a kernel size and strides of 5, with 200 output filters.

My question is around pooling, should I set the strides to equal 5 too, so that I'm not overlapping my features? And if I set the pooling size to 5, I assume I will have 1 pool per feature per filter? And if I set pool size to 2, then I will have 3 pools and here I should enable padding to equal the same?

Then on my next layer I'll use a kernel and strides of the same size as the number of pools to the previous.

Have I understood this correctly?


Your Answer

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

Browse other questions tagged or ask your own question.