2
$\begingroup$

This maybe a tough and confusing (the subject itself is confusing) question.

I understand how forward pass works in a typical multi-layer CNN (with multiple convolution, pooling, and ReLU). How does the backward pass convolution work in CNN backpropagation? That is, loss is first calculated in the output layer and how does it (the loss) backpropagate through a convolution layer?

$\endgroup$
1

1 Answer 1

1
$\begingroup$

I agree Backpropagation in CNN is difficult and generally people use frameworks for CNN without even knowing the working. You can get intuition of Backprop from this blog. The author has also given reference to several other blogs to delve deeper into its mathematics.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.