1
$\begingroup$

I read so many explanations about convolutional networks, but they seem to miss important details.

Which is true?

  1. They do backpropagation from top to bottom, down to the deepest convolutional layer.

-- or --

  1. The convolutional layers use unsupervised learning. Only the topmost layer uses supervised learning and gradient descent.

-- or --

  1. They train one convolutional layer with backpropagation to minimize the loss function. Then they use it as input, and train the second convolutional layer with backpropagation to minimize the loss function, and so on.

If (1) is true, how do they apply backpropagation to pooling layers?

If (2), what kind of unsupervised learning is used?

If you don't really know the answer, please PLEASE don't answer.

$\endgroup$

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.