1
$\begingroup$

I am trying to interpret the CNN model from the below settings. AS I am new to deep learning and I am not able to fully comprehend the layer structure . Could someone please tell me is these two settings are similar because I read in one research paper they used first setting and wrote that is similar to the second setting

  1. A CNN for MNIST with two 5x5 convolution layers (the first with 32 channels, the second with 64, each followed with 2x2 max pooling), a fully connected layer with 512 units and ReLu activation, and a final softmax output layer (1,663,370 total parameters).

  2. The CNN for MNIST has 8 layers with the following structure: 3×3×32 Convolutional → 3×3×64 Convolutional → 2×2 MaxPool → Dropout → Flatten → 1 × 128 Full connected → Dropout → 128 × 10 Fully connected → Softmax.

$\endgroup$
3
  • $\begingroup$ Can you give a link (or reference) to the paper. $\endgroup$ – Darren Cook Mar 15 at 17:11
  • $\begingroup$ Two 3x3 convolutional layers can have the same "reach" as one 5x5 convolutional layer, and uses less weights, so is a design decision that is sometimes made. But, unless you have misquoted it, that doesn't sound like the case here. $\endgroup$ – Darren Cook Mar 15 at 17:17
  • $\begingroup$ arxiv.org/pdf/2002.09699.pdf check in section V they mentioned that they are using same setting as previous paper and they defined structure in the fotter on page 7 below section V $\endgroup$ – Alex Mar 15 at 23:20
0
$\begingroup$
  1. The CNN for MNIST: 5x5x32 Convolutional -> 2x2 MaxPool -> 5x5x64 Convolutional -> 2x2 MaxPool -> Flatten -> 1x512 Full connected -> ReLu -> 512 x 10 Full connected -> Softmax.
  2. The CNN for MNIST has 8 layers with the following structure: 3×3×32 Convolutional → 3×3×64 Convolutional → 2×2 MaxPool → Dropout → Flatten → 1 × 128 Full connected → Dropout → 128 × 10 Fully connected → Softmax.

Converted to both to same formats. They are not the same.

$\endgroup$

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.