1
$\begingroup$

I have problem in understanding the equation and theory of Convolution layer from this [paper](http://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=7486259} It is said in the paper that:

enter image description here

Anyone, please help me to understand the theory and equation give.

I have rephrased that with my understanding. Please let me know if it's not correct or missed anything. enter image description here

$\endgroup$
  • $\begingroup$ Check this site cs231n.github.io/convolutional-networks and wikipedia for clearer explanations. They have some figures that make that paragraph much easier to understand. $\endgroup$ – jpmuc Mar 12 '17 at 7:24
  • $\begingroup$ Pls can you check the equation is it correct? Yes i followed these tutorials just stuck with the equation $\endgroup$ – Aadnan Farooq A Mar 12 '17 at 9:25
  • $\begingroup$ not exactly. The output has either the same size of the input (with zero padding, in your case N x N) or like in the text you quote, some other size, depending on the size of the convolution kernel, but not k x k. Please go carefully through the link I provided you. $\endgroup$ – jpmuc Mar 14 '17 at 20:59
  • $\begingroup$ It may seem unrelated, but when implementing CNN, this link helped me understand it better, and also how to implement convolutions efficiently (for CNNs). petewarden.com/2015/04/20/… $\endgroup$ – jpmuc Mar 14 '17 at 21:00

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.