Nice post.



Nice post.

But, in CONV2, why the

NUMBER OF PARAMETERS

is

(5 * 5 * 96 + 1) * 256=614656

The filter size in CONV2 is 5*5*48 (Data from AlexNet paper). So, the number of pareameters will be (5*5*48+1)*256.

Is that right?

Source: Deep Learning on Medium