In what scenario when assembling a DL CNN would you want to have two adjacent pooling layers, without a convolutional layer between?

  • $\begingroup$ Why cannot 2 single pooling layers converted to a single pooling layers? $\endgroup$ – DuttaA Jul 30 '18 at 16:44
  • $\begingroup$ I guess that's a part of the question. Would anyone ever use two layers? I know that CNTK won't complain if you build it like that but is there ever a reason to? $\endgroup$ – Gaius Jul 30 '18 at 20:03
  • $\begingroup$ To the best of my knowledge, I did not find any situation where I had to use 2 adjacent layers. If you have seen it somewhere please share the source or is it a hypothetical question? $\endgroup$ – varsh Jul 31 '18 at 16:22
  • $\begingroup$ I was asked the question was it possible and yes it is possible in the sense that you could write the code that way and it would run. However I answered no because I couldn't think of a situation in which you would, and that was the wrong answer. So I have come to the community for some guidance and insight. $\endgroup$ – Gaius Jul 31 '18 at 19:25

In your question you didn’t specify the type of pooling that you aren’t doing. So it’s possible that you could have, for example a mean pool followed by a max or min pool. What this could do is combine the idea of reducing dimensionality of your data from a holistic perspective with the mean pool and then choosing the best of your averages with your max pool.


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.