Hi, I hope everybody is doing well.
I think there are wrong dimensions given by the grader in the second assignment of week one, the function name is convolution_model.
I think it should be filter_size3number_of_filter+biases, and correct value should be according to that formula( 223*16+16 = 208) but it is written 528, can anybody explain why?
Hi, I have looked into the problem you are referring.
It is a CONV layer with filter size 2 * 2 * 8 and stride of 1 , padding is SAME and there are total 16 filters, ( previous layer output is of shape 64 * 64 * 8 )
One thing to note is that for each filter there is only one bias ,
so,
No of params = (# of params in one filter + 1 ) * # of filters
= (2 * 2 * 8 + 1 ) * 16
= 528
(note :- we added 1 for the bias term)